Light field image feature point detection method based on multi-scale Harris

Abstract

The invention discloses a light field image feature detection method based on multi-scale Harris. The method specifically comprises the steps: reading a light field original image parameter file intoMATLAB, and decoding and processing the light field original image parameter file into an effective four-dimensional light field matrix; taking the maximum value in the angle resolution [u, v] as thelength n of a slope list to obtain the slope list, and refocusing each slope in the slope list to obtain a focus stack image corresponding to the slope; carrying out multi-scale Harris corner detection on each focus stack image; carrying out non-maximum suppression on the corner detected in each scale of the current corner stack, and reducing the influence of multiple responses; carrying out multi-scale judgment on the candidate corner points; if the candidate corner points appear in multiple scales, reserving the candidate corner points; otherwise, deleting, wherein the finally reserved angular points are the feature points of the light field image; obtaining the real information of the whole space by adopting the position information and the angle information in the light field image. Therefore, the defects of shielding, depth loss and the like of the traditional imaging are overcome, and the scene description is more comprehensive.

Description

technical field [0001] The invention belongs to the technical field of digital image processing, and in particular relates to a multi-scale Harris-based light field image feature point detection method. Background technique [0002] There are always some unique pixel points in the image, which can be regarded as the features of the image, called feature points. Many applications in the field of computer vision are based on image feature points. Therefore, the definition and detection of image feature points becomes particularly important. In the field of computer vision, the detection of image feature points has been widely used, such as image matching, image stitching, fault detection and 3D reconstruction, etc. [0003] Image feature point detection is the premise of image analysis and recognition. It is the most effective way to simplify and express high-dimensional image data. It is difficult to directly extract useful information from the data matrix of an image, so it...

AI Technical Summary

Problems solved by technology

Moreover, the Harris corner detection operator is not sensitive to brightness and contrast changes, and also has rotation invariance, but not scale invariance.

Images