Abstract: Object detection approaches need to perform accuratelyand robustly over a wide range of scenes. It would bequite valuable if one can devise a performance indexfor an object detection approach as a function of thenature of a particular scene. Basically this requiresan ability to derive a quantitative measure for the'clutter' observed in an image. Most images of interestare texture-rich i.e. the important perceptualproperties are based upon the spatial arrangements ofsimple patterns which might be regular in nature. As aresult, it is natural to utilize texture analysis basedoperators to define the measure of image quality of'clutter' that is being sought. It has been proven thatthe gray level cooccurence (GLC) matrices of an imageembody important texture information, and the image canindeed be reconstructed from these matrices. Hence itis proposed that GLC-based measures be derived and usedto quantify image quality. Current approaches are basedon only one of several important perceptuallymeaningful measures which can be computed from GLCmatrices. Prior work done in this area is assessed inthis paper. The derivation of the image qualitymeasures from GLC matrices is currently beingresearched. This paper presents a discussion of theseissues along with the objectives and results of anongoing study involving object detection in highresolution aerial images.!
展开▼
