Comparative study on image segmentation techniques
- 1. Submitted by, G. Midhu Bala and J.Asenath
- 2. Introduction Image processing is any form of signal processing for which the input is an image, such as a photography or video frame. The output of image processing may be either an image or a set of characteristics related to the image. Image Analysis - to extract high level information on an image. Image Segmentation - to change the representation of an original image into meaningful portions which makes it easier to analysis. To locate objects and boundaries.
- 3. Segmentation Techniques •Thresholding •Edge Detection •Color Image Segmentation •Histogram Based Method
- 4. Types of images 8 bit image 8 bit image RGB
- 5. 16 bit image 16 bit RGB
- 6. 32 bit image 32 bit RGB
- 7. 8 bit Color 8 bit color RGB
- 8. RGB Color
- 9. RGB Stack Red Green Blue
- 10. File formats JPG uses lossy compression GIF always uses lossless LZW compression, but it is always an indexed color file (8-bits, 256 colors maximum), which is poor for 24-bit color photos. PNG is transparency for 24 bit RGB images. lossless compression, of different types), but PNG is perhaps slightly slower to read or write. TIF is lossless (including LZW compression option), which is considered the highest quality format for commercial work.
- 11. Test image . Original disease free leaf Original affected leaf
- 12. Thresholding Original image into binary image Foreground can be separated from the background by selecting the threshold value Global Thresholding -only one threshold value for entire image Local Thresholding - different value for different regions Methods Edge Based - detects and links edge pixels to form contour. Region Based - detects the entire region
- 13. Threshold Value : 100 Threshold Value : 150 Threshold Value : 255 Global Thresholding
- 14. Otsu Method Threshold Value : 193
- 15. Edge Detection Reduce the amount of data in an image. Provides ability to extract the exact edge. Corners, lines, curves . Meaningful discontinuities in the grey level. Edge detected image
- 16. Canny Edge Detection:(Criteria) Detection: The probability of detecting real edge points is maximized and falsely detecting non-edge points is minimized. This corresponds to maximizing the signal-to-noise ratio. Localization: The detected edges should be as close as possible to the real edges. Number of Responses: One real edge should be result in more than one detected edge.
- 17. Canny Edge Detection Algorithm Smoothing: Blurring of the image to remove noise. Finding gradients: The edges should be marked where the gradients of the image has large magnitudes. Non-maximum suppression: Only local maxima should be marked as edges. Double thresholding : Potential edges are determines by thresholding. Edge tracking by hysteresis: Final edges are determined by suppressing all edges that are not connected to a very certain strong edges.
- 18. Color Image Segmentation: Color image segmentation is used to extract high level information of the image based on color. Three phases are Phase1: Preprocessing: Morphological methods are applied to remove the noises away from image which applied to smooth some spots on uniformed patterns. Phase2: Transformation: Color space transformed methods are used to transform other color space to RGB.
- 19. Phase3: Segmentation: Applying clustering algorithm like K-means algorithm for finding the appropriate cluster numbers and segment images in different color spaces. The cluster with the maximum average variance is split into new clusters. Segmented image
- 20. Histogram- based methods: Compute- Pixels , peaks, valleys Locate – clusters Recursively applied for finding the smaller clusters. Distinguishes the two homogeneous regions of the foreground and background of an image. Histogram
- 21. Conclusion Partitioning an Image using segmentation techniques leads to extract different regions with similar attributes . It also detects high level information of an image for image analysis and further researches.