Keypoint-based Concise Representation Of Image Information And Its Appliactions

Visual sense is the main source for human to obtain information. But the visual image information is very large, so if the information can be represented concisely,that will greatly improve the efficiency of the image analysis.With the hardwork of researchers many concise representation of image information methods were proposed such as methods based on projection, methods based on compressing and methods based on keypoint detection. However, those methods have many drawbacks. So this topic on one hand consummate the existing concise representation of image information methods and on the other hand, the existing methods are applied to new directions.First, to overcome the drawback of conventional histogram, a novel histogram construction method has been proposed based on keypoint detection. The importance of gray levels in different positions can be calculated by using the scale of keypoints. Meanwhile, the statistic results are weighted according to the importance of every pixel keypoint. The new histogram method could reflect the influence of the gray levels on the image information objectively. Then the novel histogram is used to an image enhancement method based on histogram equalization. Result shows that this image enhancement algorithm could enhance the important information in information area, and inhibit the noise in background area.Most of the subjective assessment methods are tedious and those objective assessment methods are inaccurate. So, a simple objective quality assessment based on keypoint detection for enhanced images is proposed. Test results on traditional image enhancement algorithms show that this model’s predictions agree well with subjective assessment.SIFT(Scale-Invariant Feature Transform) method is used in suspected gas source localization, SIFT features are invariant to scale and to image scaling and rotation, and partially invariant to orientation change in illumination and 3D camera view points, so SIFT is better in object matching, and the experiment result shows that this method is advanced and effective...