A Statistical Model for Shadow Removal of ManMade Objects and Change Detection in Satellite Images
Kesini Krishnan V1, Smitha P. S2

1Kesini Krishnan V, M.Tech Student, SCT College of Engineering, Pappanamcode, Trivandrum (Kerala), India.
2Smitha P. S, Asst. Prof., SCT College of Engineering, Pappanamcode, Trivandrum (Kerala), India.

Manuscript received on 15 June 2015 | Revised Manuscript received on 25 June 2015 | Manuscript Published on 30 June 2015 | PP: 147-152 | Volume-4 Issue-5, June 2015 | Retrieval Number: E4116064515/15©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: In this paper shadow detection and removal done as the pre processing steps for change detection beacause the presence of shadow causes mistakes in change map. For shadow detection there is a convexity analysis which are multiphase object segmentation and thresholding for suspected and false shadow removal also considering the object properties such as shape ,area ,perimeter ,average gray scale value and standard devitaion for more perfection. Shadow removal is employed by the method IOOPL (Inner Outer Outline Profile Lines) matching and relative radiometric correction. For IOOPL generation, first forms the object boundaries,then form two additional boundaries by expanding and contracting object boundaries. Build a graph,grayscale versus no of points. When doing a similarity test in IOOPL graphs secton by section, matching coefficient become high, that region treated as homogeneos and data reconstructed compared to non shadow area. Removal is done by relative radio metric correction. In Change analysis find the binary descriptors of each pixel in the two images and find hamming distance as similarity measure between binary descriptors of each pixel at the same location in two images. After there is a ranking system in change analysis and which is done by Lloyd-Max Quantization. Here in this paper we employed M=2, M=3 quantization levels In shadow treatment validation is done by comparing the gray scale average and standard deviation of non shadow area,shadow area and shadow removed area. The results are shown that it is very efficient compared to existing methods. . Shadow detection and removal is 93% accurate compared to existing methods. Average running time of change detection is better compared to previous works. Also most of the previous works are dealing with two levl quantization. Here more than 2 levels can be done with in seconds.
Keywords: Multiphase Segmentation, Histogram, Thresholding, Inner Outer Outline Profile Lines (IOOPL), Shadow Detection, Shadow Removal, Change Detection, Binary Descriptor, Hamming Distance, Lloyd Max Quantization

Scope of the Article: Image Processing