Color information is of paramount importance used by computer vision systems in various fields like image understanding and pattern recognition. However, the quality of information is degraded by certain types of noise during image acquisition and transmission. Impulse noise is one such noise which occurs for a short duration. Presence of impulse noise complicates the subsequent stages of image processing such as edge detection, segmentation, etc. Hence, noise filtering is the most important task in signal processing. VectorMedianFilters (VMF) (Astola et al., 1900), Basic Vector Directional Filters (BVDF) (Trahanias and Venetsanopoulos, 1993) and Directional Distance Filter (DDF) (Karakos and Trahanias, 1995) are the most common filters used for removing impulse noise. These are nonlinear filters based on order statistic. But they have the drawback of excessive blurring leading to loss of fine details. This is because they filter every pixel without checking the presence of an impulse. In order to improve the performance of VMF and its extensions, adaptive switching filters are developed. Switching filters detect the presence of an impulse and filtering is performed if found corrupted. Adaptive Center-Weighted VMF (ACWVMF), Adaptive Center-Weighted VDF (ACWVDF) and Adaptive Center Weighted DDF (ACWDDF) (Lukac and Smolka, 2003;
This paper introduces a class of nonlinear multichannel filters capable of removing impulsive noise in color images. The here- proposed generalized selection weighted vector filter class constitutes a powerful filtering framework for multichannel signal pro- cessing. Previously defined multichannel filters such as vectormedian filter, basic vector directional filter, directional-distance filter, weighted vectormedianfilters, and weighted vector directional filters are treated from a global viewpoint using the pro- posed framework. Robust order-statistic concepts and increased degree of freedom in filter design make the proposed method attractive for a variety of applications. Introduced multichannel sigmoidal adaptation of the filter parameters and its modifica- tions allow to accommodate the filter parameters to varying signal and noise statistics. Simulation studies reported in this paper indicate that the proposed filter class is computationally attractive, yields excellent performance, and is able to preserve fine details and color information while eﬃciently suppressing impulsive noise. This paper is an extended version of the paper by Lukac et al. presented at the 2003 IEEE-EURASIP Workshop on Nonlinear Signal and Image Processing (NSIP ’03) in Grado, Italy.
ervation but having the same MSE and the same MAE with respect to the original picture. To obtain this result, we considered a 512 × 512 version of the 24-bit color image “Parrots” and two filters with different window sizes, such as the 5-point and the 5 × 5 vectormedianfilters . We generated the noisy input picture as fol- lows. We considered versions of the “Parrots” image corrupted by different densities of noisy pixels, i.e. pixels where each channel component is incremented (or dec- remented) by a fixed value. We searched for the noise density p and the noise amplitude d such that these me- dians yield filtered images with the same RMSE and the same MAE. A satisfactory result was obtained by choos- ing p = 39% and d = 37. Clearly, the 5 × 5 operator (Figure 1(b)) is more effective than the 5-point filter (Figure 1(a)) in removing noise at the price of a worse detail blur, but these very different features are not ap- parent from RMSE and MAE measurements that yield RMSE = 8.8 and MAE = 3.6 for both filtered images.
Over the last decade, several filtering algorithms have been proposed to overcome the problem of random impulse noise in color images. The vectormedian ﬁlter (VMF) is one of the most effective with regard to noise suppression and computing efficiency. Unfortunately, when the noise ratio is high, some edges and other details can be destroyed by a traditional vectormedian filter. Various modifications to medianfilters have been proposed over the last three decades to deal with this problem. Trahanias and Venetsanopoulos  proposed a basic vector directional filter (BVDF) and Karakos and Trahanias  presented a directional-distance ﬁlter (DDF). BVDF and DDF enable the separate processing of vector-valued signals via directional processing in order to maintain image sharpness. This approach has proven successful in dealing with noise of high density.
Because different groups of pigeons were tested, we were able to pool the data collected at the two release sites according to their deviation from the home directions and by setting the home direction to North (Table 1, Fig. 1). The pooled distributions of groups C and V1 were significantly different from uniform according to both the Rayleigh and V-test (see Table 1) and their vectors were homeward directed (see the confidence limits given in Fig. 1). By contrast, the pooled distribution of the ON birds was significantly different from uniform according to the Rayleigh test, but not according to the V-test (see Table 1), which takes into account the expected direction. In fact the ON pigeons mean vector was directed opposite to the home direction (see the confidence limits given in Fig. 1). The three pooled vanishing distributions were not statistically different in dispersion (Kruskal–Wallis, P >0.1), but they were significantly different in orientation (Kruskal–Wallis, P <0.001). In
In this paper we present an architecture based on local segmentation process for salt-and-pepper noise reduction by using median filter. The approach is implemented efficiently in hardware. It is based on partitioning of an image matrix in homogeneous mask. After that with the help of median filter we get the median values. These median values are different from the actual values. The size of these values (pixels) is nearly of the same as that of the remaining elements of the matrix. Now, these values are passed to a 4×1 MUX, with the help of MUX operation one can take single output at a time which is used to set the threshold o f an image. The motivation of our design is to gets good visual standard of an image.
The estimation of RTO based on median filter algorithm is proposed and analyzed in this paper. Median filter is a useful filter due to its preserving characteristic for image processing applications. From the experimental results, we find that it per- forms well for heavy-tailed distributions to eliminate impulses in bursty traffic flows. As a result, median filter can perform better than the weighted-average filter because consistent RTT and small RTO are obtained, which are desirable factors for high connection throughput to alleviate traffic congestion.
Image processing is an important area in the information industry.  Image de-noising has been one of the most important and widely studied problems in image processing and computer vision. It refers to the task of recovering a good estimate of the true image from a degraded observation without altering and changing useful structure in the image such as discontinuities and edges. Impulse noise is one of very common and intensive noise, which effects the images during the acquisition, transmission and processing of images. Impulsive noise can be Salt & Pepper Noise (SPN) or Random Valued Impulsive Noise (RVIN).The need to have a very good image quality is increasingly required with the advent of the new technologies in a various areas such as multimedia, medical image analysis, aerospace, video systems and others. Obviously it is impossible that there is no noise in the image. So it is important to eliminate the noise before the further process. We can first apply the median filter (MF) to the corrupted image. The median filter is by far the most useful order-statistic filter .Nevertheless, to one’s disappointment is removing fine details tendency for MF.
Abstract— Many techniques continue to be proposed so far to get rid of the noise through digital images with more optimistic method. Each technique has its very own drawbacks. Although Fuzzy Mean Median (FMM) has demonstrated promising results on the available techniques, given it utilizes the top features of data mining to get rid of mixed noise. This data mining method is employed to check which sort of noise occurs in the picture. But it has not yet considered the enhancements with the filtering techniques i. e. improved bilateral filtration system and decision based alpha trimmed median filtration system. Therefore this perform has proposed some sort of novel data mining and improved bilateral and also decision based alpha trimmed based filtration system. The comparison offers clearly shown that the proposed technique outperforms on the available one.
Mean Filter – This is a linear type spatial filter where a specified matrix is used as kernel or window to mask each image pixel. Based on the average of the kernel matrix and the neighboring pixel values, the resultant pixel is derived for which the filtered image has a smooth effect  and it works well against grain noise , which is also known as film grain. With the increasing kernel size, the strength of averaging will be amplified and as a result the image will be blurred. Due to its linear characteristic, it is not able to preserve the edge. Moreover, the Peak Signal to Noise Ratio (PSNR) is lower than that obtained by non-liner filters .
noise in color images. On the other hand, the standard median filter  or its multi-channel extensions, i.e., the vectormedian filter  and the basic vector directional filter , are unable to adapt their behavior to varying noise and signal statistics related to the local image information of the samples inside a sliding filtering window. These filters performing the fixed amount of smoothing result in blurring of fine image details.
Digital image analysis plays a vital role in medical imaging like magnetic resonance imaging, ultra sound imaging, X-ray and computed tomography. Departure of the ideal signal is usually referred to as noise. Noises in such digital images arise during image acquisition and/or transmission. The data dropout noise is generally called as speckle noise. Speckle noise is a multiplicative noise that degrades the visual evaluation in medical imaging. Speckle noise suppression plays a very essential role in diagnosis. The image acquisition devices need despeckling techniques for medical imaging in routine clinical practice. Image filtering is an important technique used for the detection and removal of noise from the digital images. Median filter has been introduced by Turkey in 1970 . It is a non-linear filter used for smoothing the images. Sudha et al  recommends a novel thresholding algorithm for denoising speckle noise in ultrasound images with wavelets. An improved adaptive median filtering method for denoising impulse noise was carried out by Mamta Juneja et al . Thangavel et al  showed that the M3-filter had performed better than Mean, Median, Max, Min and various other filters. The Hybrid Max filter which performs significantly better than many other existing techniques for removal of speckle noise was shown by Gnanambal Ilango et al . R. Marudhachalam and Gnanambal Ilango  proposed different Center Weighted Hybrid filtering techniques for denoising of medical images. The objective of this study is to develop new filtering techniques based on different metric topological neighbourhoods and investigate their performance on CT
The high mortality rate in women due to breast cancer prompts the need for early detection techniques. Towards this, mammography imaging system is an advantage for both early detection and screening of breast cancer tumors, given that it uses low dose x-rays for examining the breasts. The use of screening mammography in the early on detection of breast cancer tumor of variable sizes (malignant or benign) reduces mortality rate in women. The objective of this study is to enhance the current accuracy (diagnostic) of digital mammograms using industry standard simulation software tool, MATLAB and the MIAS dataset. The technique involves the identification of tumor cells in terms of different stages of the disease. The processes of recognition and classification of mammograms were concisely analyzed with the aim of differentiating between normal and abnormal (benign or cancerous). It is reported that dense breasts can make the interpretation of conventional mammograms more complex. Although advanced digital mammography techniques assert better detection in dense breast tissues, the availability of such costly digital mammograms is not widespread. This problem can be minimized by analyzing different breast structures (mammograms) using the MATLAB numerical analysis software for image processing applications. In this study, we investigate wavelet-based feature extraction from mammogram images and efficient dimensionality reduction technique. The objective is to propose a new computerized feature extraction technique to identify abnormalities in the mammogram images. In this work feature reduction is carried out using median idea for both maximums and minimums high frequency coefficients sub bands. The proposed system was tested on the MIAS database, resulting in 98.43% succession rate of normal and abnormal classification while for benign and malignant classification, 98.26% rate was obtained.
Noise removal plays vital role in image processing and also important pre processing task before performing post operation like Image segmentation etc.. This paper presents a effective and efficient algorithm in order to remove impulse noise from gray scale and color images. Challenging results show the superior performance of the proposed filtering algorithm compared to the other standard algorithms such as Standard Median Filter (SMF), Median Filter (MF), Weighted Median Filter (WMF) and Trimmed Median Filter (TMF). Furthermore, various performance metrics such as the MSE, PSNR and SSIM have been compared with Existing standard algorithms. The computational time for the denoised image is calculated for different noise levels and the proposed algorithm has lower computational time, hardware complexity and ease in operation.The obtained results prove that it has better qualitative analysis by improving visual appearance and challenging quantitative measures even at high noise densities ranging up to 90%.
There are many filters for speckle reductions with better visual interpretations, while other good noise suppression or smoothing capabilities. Some of the best known speckle reduction filters are Lee Kuan, median, standard Frost, Enhanced Frost, Weiner, Gamma MAP and SRAD filter. Some of these filters have unique speckle reduction approach, spatial filtering in a square - move window as known kernel performs. The filtering is based on the statistical relationship between the center pixel and the surrounding pixels are calculated. The typical size of the filter window can range from 3-mal 3-33 -by -33 rich, but the size of the window to be odd. If the size of the filter window is too large, important information will be lost by over-smoothing. On the other hand, if the window size is too small, speckle reduction may not produce good results. In general, a 3-by-3-or 7- used by -7 windows is accepted to give good results .
The multi-resolution images are segmented using Support Vector Machine Classifier. The proposed method was analyzed based on Sensitivity (99.4%), Specificity (99.6%), Positive Predictivity (97.03%) and Accuracy (99.5%) respectively. The segmented brain tumor images were compared with the ground truth images to prove its superiority or accuracy. Even though the authors claimed for higher tumor segmentation accuracy, they failed to specify the datasets used and the number of images used for evaluation.
In this paper, our object is to get sharp edges of an image in the presence of the salt-and-pepper noise. To remove this noise we use median filter. We detect edges of this output with the help of classical edge detectors i.e. LoG operator, sobel operator, prewitt operator, but the edges we get are not as sharp. So, we apply our proposed method. We apply adaptive histogram equalization on the output of median filter and perform filtering. Adaptive histogram equalization enhances the contrast of the gray scale image by transforming the values using contrast-limited adaptive histogram equalization. It operates on small regions in the image, called tiles, rather than the entire image. Each tile's contrast is enhanced. The contrast, especially in homogeneous areas, can be limited to avoid amplifying any noise that might be present in the image. After it we detect their edges with the help of given edge detectors. Thus, we get sharp edges of de-noised image. Proposed algorithm of this paper is shown in figure below.
Image classification and segmentation are the two main important parts in the 3D vision system of a harvesting robot. Regarding the first part, the vision system aids in the real time identification of contaminated areas of the farm based on the damage identified using the robot’s camera. To solve the problem of identification, a fast and non-destructive method, Support Vector Machine (SVM), is applied to improve the recognition accuracy and efficiency of the robot. Initially, a median filter is applied to remove the inherent noise in the colored image. SIFT features of the image are then extracted and computed forming a vector, which is then quantized into visual words. Finally, the histogram of the frequency of each element in the visual vocabulary is created and fed into an SVM classifier, which categorizes the mushrooms as either class one or class two. Our preliminary results for image classification were promising and the experiments carried out on the data set highlight fast computation time and a high rate of accuracy, reaching over 90% using this method, which can be employed in real life scenario.
Abstract— In digital image processing, scanned images are degraded by various kinds of noise. When an image is transformed from one form to another during scanning, transmitting, digitizing, storing etc., degradation occurs in the output image. Hence, the output image needs to be enhanced in order to be better analyzed. Denoising is considered to be one of the effective pre-processing techniques in digital image processing. This paper investigates the performance of linear as well as non linear filters namely; Gaussian filter, Mean filter, Median filter, Sobel filter and Canny filter for removing several noises such as Gaussian noise, Salt & Pepper noise, Speckle noise and Poisson noise. The performance of the above five linear and non linear filters is compared by using two performance parameters namely; Mean Square Error (MSE) and Peak to Signal Noise Ratio (PSNR). The experimental result shows that median filter performs best than other five filters for both performance parameters.
attention model considering the energy of motion vector and the spatio-temporal correlation to analyze motion attention on the basis of analysis of the motion vector. Guironnet and Zhai  proposed an attention model based on spatio-temporal information fusing static and moving target model in 2005. Jing Zhang  and Se- ung-Hyun Lee  applied extraction of region of atten- tion (ROA) to target segment on static image and achieved much effects. Junwei Han  took advantage of attention model to segment video target. The global motion estimation and compensation was used and static attention and dynamic attention fusion was carried out to get the final result, but this method is limited to the local motion scene.