240 373 Image Processing Montri Karnjanadecha montricoe psu
- Slides: 19
240 -373 Image Processing Montri Karnjanadecha montri@coe. psu. ac. th http: //fivedots. coe. psu. ac. th/~montri 240 -373: Chapter 9: Image Segmentation 1
Chapter 9 Image Segmentation 240 -373: Chapter 9: Image Segmentation 2
Image Segmentation • Segmentation – Splitting an image up into segments (area or regions( – Each segment holds some property distinct from its neighbor – Basic requirement for identification and classification of objects in a scene – Answering questions like: How many object are there? Where is the background? How large is the object? 240 -373: Chapter 9: Image Segmentation 3
Image Segmentation • Segmentation can be approached from two points of view. 1 by identifying edges (or lines) that run through an image ( edge operation(. 2 by identifying regions (or areas) within an image ) region operation( Note: 1 and 2 are correlate: completion of an edge detection is equivalent to breaking one region into two 240 -373: Chapter 9: Image Segmentation 4
Region Operations • Required to cover a substantial area of the scene rather than a small group of pixels 240 -373: Chapter 9: Image Segmentation 5
Region Operations Technique 1: Crude region detection USE: To reconsider an image as a set of regions. OPERATION: The regions are simply identified as continuous pixels of the same gray level. The boundaries of the regions are at the cracks between the pixels rather than at pixel positions. 240 -373: Chapter 9: Image Segmentation 6
Region Operations Notes • This technique may give too many regions to be useful • Bunching (quantizing) technique can be used to reduce number of regions 240 -373: Chapter 9: Image Segmentation 7
Region Merging Technique 2: Region merging USE: To reduce the number of regions, to combine fragment regions, to determine which regions are really part of the same area. 240 -373: Chapter 9: Image Segmentation 8
Region Merging OPERATION : – Let s be a crack difference, i. e. the absolute difference in gray levels between two adjacent (above, below, left, or right) pixels – Then given a threshold value, T, we can identify for each crack i. e. w=1 if the crack is below the threshold (the regions are likely to be the same), or 0 if it is above threshold. 240 -373: Chapter 9: Image Segmentation 9
Region Merging – Measure the full length of the boundary of each of the regions that meets at the crack. These will be b 1 and b 2. – Sum the w’s that are along the length of the crack between the region and calculate – If this is greater than a further threshold, deduce that the two regions should be joined 240 -373: Chapter 9: Image Segmentation 10
Region merging example If we make T=3, then all the cracks are significant 240 -373: Chapter 9: Image Segmentation 11
Region merging example Region A B C Boundary 17 10 4 240 -373: Chapter 9: Image Segmentation Sum of w’s With A With B 9 0 9 4 With C 0 4 - 12
Giving sum/min(b 1, b 2) for A, B, and C as follows: Region A B C A 0. 9 0 240 -373: Chapter 9: Image Segmentation B 0. 9 1. 0 C 0 1. 0 - 13
– If we take threshold for combining as 0. 95, then B will be combined with C. – If the combining threshold is set lower, then A will also be combined giving the whole image as just one region. • The algorithm combines the small region (4) with the larger region (2), rather than the two larger regions (1 -2. ( • Size of the region can be used instead of the length of the region boundary. 240 -373: Chapter 9: Image Segmentation 14
Region Splitting Technique 3: Region splitting USE: To subdivide part of an image into regions of similar type OPERATION : – Identify significant peaks in the gray-level histogram – Look in the valleys between the peaks for possible threshold values. – Find splits between the best peaks first 240 -373: Chapter 9: Image Segmentation 15
Region Splitting LIMITATIONS: – This technique relies on the overall histogram (Multiple chessboards example( – Example 240 -373: Chapter 9: Image Segmentation 16
Region Splitting *********** 0 **** 1 the valley is at 1 ***** 2 ********* 3 **** 4 240 -373: Chapter 9: Image Segmentation 17
Region Splitting – Which side should the 1’s go? – Consider both options (2, 3, 4) , (0, 1)(1, 2, 3, 4) , (0) OOOOOO. O. . OOO. . . . O. O. . . . OO. . . . OOOOOO. O. OOOOOOO. O. . . . OOO. . . . The second option less stragglers than the first but is is difficult to see this without trying it. 240 -373: Chapter 9: Image Segmentation 18
Region Splitting COMMENTS: – The histogram could just be of a region within the image – For color images, there are 3 histograms to choose from 240 -373: Chapter 9: Image Segmentation 19
- Neighborhood processing in image processing
- Point processing in image processing
- Histogram processing in digital image processing
- Nonlinear image processing
- Point processing in image processing
- Gonzalez
- Image transform in digital image processing
- What is image restoration in digital image processing
- Spatial and temporal redundancy in digital image processing
- Key stages in digital image processing
- Huffman coding example
- Image sharpening in digital image processing
- Geometric transformation in digital image processing
- Digital image processing diagram
- Image transform in digital image processing
- Maketform matlab
- Noise
- Eecs373
- Eecs 511
- Eecs 373