SCK: A Sparse Coding Based Key-Point Detector

Abstract

All current popular hand-crafted key-point detectors such as Harris corner, MSER, SIFT, SURF… rely on some specific pre-designed structures for the detection of corners, blobs, or junctions in an image. In this paper, a novel sparse coding based key-point detector which requires no particular predesigned structures is presented. The key-point detector is based on measuring the complexity level of each block in an image to decide where a key-point should be. The complexity level of a block is defined as the total number of non-zero components of a sparse representation of that block. Generally, a block constructed with more components is more complex and has greater potential to be a good keypoint. Experimental results on Webcam and EF datasets [1,2] show that the proposed detector achieves significantly high repeatability compared to hand-crafted features, and even outperforms the matching scores of the state-of-the-art learning based detector.

Image Feature Information Extraction for Interest Point Detection: A Comprehensive Review

“…The first-and secondorder horizontal and vertical directional derivatives are used to train a detector according to the criterion that the detected interest points under image affine transformations should have a high detection repeatability. Hong-Phuoc et al [215] applied a sparse coding technique to analyze the complexity of interest points on images, and an image pixel where the information complexity reaches a certain level is determined as an interest point.…”

A Scale and Rotational Invariant Key-point Detector based on Sparse Coding

“…The algorithm for detection of key-points in a gray-scale image using SS-SCK [9,10] is briefly reviewed below with justification for SS-SCK's adaptability to different contexts provided in [9].…”

“…From ( 8) to (10), we can calculate a, b, c of the model. The row coordinate of the maxima or the key-point in sub-pixel accuracy is then located at where the derivative of the parabola equals to zero.…”

“…As will be demonstrated later in the experiments, SRI-SCK is very robust against significant changes in illumination as its precursors [9,10]. This is because for each block 𝑿 = 𝑿(𝑥, 𝑦, 𝑙), the input to the sparse coding algorithm is not its original, but the normalized version 𝒀 𝒏𝒐𝒓𝒎 ′ of its circular masked version Y after being reshaped to a vector 𝒀 ′ .…”

“…Recently, a Sparse Coding based Key-point detector (SCK) which relies on no pre-designed structures was proposed [9,10]. As being independent of any human designed structure, SCK is very flexible to context changes.…”