NOISE FILTERS IN IMAGE PROCESSING

1. LINEAR FILTERS
2. NON-LINEAR FILTERS
3. SHARPENING FILTERS

1. INDIAN INSTITUTE OF FOOD PROCESSING TECHNOLOGY PRESENTED BY- ANIMESH SINGH SENGAR M.TECH. FPE 2017694604

2. Noise is an unwanted or distort signal that may corrupt the quality or the originality of the image. P roduced due to transmission U ndesirable effects such as artifacts, unrealistic edges, unseen lines, corners, blurred objects and disturbs background scenes can also be seen Noise in Image Processing

3. Continued.. Its important and necessary to remove the noise after image digitization. T wo images taken one after another in less than 2 s from the same scene, using the same image acquisition system and their differences to demonstrate the noise during the image acquisition M ost efficient and feasible approach for image noise removal is to “average” the image by itself

4. Some reasons of Image Degradation Poor image sensors Defects in optical lenses in camera Non-linearity of the electro-optical sensors Graininess of the film material which is utilized to store the image Relative motion between an object and camera Wrong focus of camera Atmospheric turbulence in remote sensing Degradation due to temperature sensitive sensors Poor light levels

5. Image subjected to different type of Noises

6. Types of Filters

7. Linear filtering technique is used for reducing random noise In this value of output pixel is a linear combination of the values of the pixels in the input pixels neighborhood Weight and size of filter can be adjusted to remove the different types of noises Increasing the size of the filter results in a smoother image with less noise , but meanwhile the details of the image are reduced . Linear Filter

8. We have to place this mask over the image and multiply each component of this mask with the corresponding value of the image and then we will take summation of all these products. Averaging Operation

9. 3 X 3 5 X 5 7 X 7 Original Processed image is blurred and the details are reduced. Averaging operation is equivalent to integration operation. So , if we integrate a image what we will get is blurred and smoothing effect.

10. Linear Filter (Types)

11. Gaussian filter Gaussian filter is used to remove Gaussian noise and will not blur the image too much, which enables Gaussian filter to perform well in edge detection Probably the most useful filter (although not the fastest). It effectively reduces the noise and image details It is being used in image processing of meat .

12. Wiener filter Wiener filter was proposed by Norbert Wiener in 1940. Its purpose is to reduce the amount of noise in a signal. The aim of the process is to minimize Mean Square Error i.e., Difference between the original signal and the new signal should be as small as possible . Power spectra of original noise must be known, before its application. If the variance is large, wiener performs little smoothing If it is small, wiener performs more smoothing

13. Median Filter (Non-linear) Intensity values of pixels in a small region with the size of the filter are examined, and the median intensity value is selected for the central pixel . Two primary advantages – does not reduce the brightness difference of images, as the intensity values of the filtered image are taken from the original one. median filter does not shift the edges of images. These are most effective against random noise or salt and pepper noise.

14. Example …

15. Standard Median Filter Filtering is conducted across to each pixel without considering the contamination status of an image. C orrupted pixels with contributions in the image details are filtered, which results in the loss of some image information M echanism detecting impulse noise is introduced before performing image filtering to solve this. detection mechanism would enable the corrupted pixels to be filtered and uncorrupted ones to be kept intact .

16. Hybrid Median Filter Modified version of median filter and used in removing impulse noise without losing edges . P ixel value of a point is replaced by the median pixel value of 4-neighborhood of the point, median pixel value of cross neighbors of the point , and m edian pixel value of the point

17. Progressive Switching Median Filter Used to detect and remove the impulse noise from highly corrupted images. Impulse pixels located in the middle of large noise blotches can be properly detected and filtered, showing better results in up to 60% noise Performance drastically reduced when the noise is over 60 %

18. Adaptive Median Filter U sed for discriminating corrupted and uncorrupted pixels N oisy pixels replaced by median value and uncorrupted pixels left unchanged. F ilter performs better at low noise densities . Better noise removal is also obtained at higher noise densities after increasing the window size.

19. Sharpening Filters We have seen, integration operation is giving blurred and smoothing effect So, its opposite differentiation function will give sharpening effect Two type of derivatives are used First order derivative Second order derivative First order derivatives generally produces thicker edges in an image Second order derivatives give stronger response to fine details such as thin lines and isolated points

20. First order derivative have stronger response to gray level step Second order derivative produce a double response at step edges Second order derivatives are better suited for image enhancement.

21. Thank you