Sketch-based image retrieval by shape points description in support regions

Abstract

We introduce a statistical shape descriptor for Sketch-Based Image Retrieval. The proposed descriptor combines feature information in near and far support regions defined for each sketch point. Two feature values are extracted from each point, corresponding to near and far support regions from the point's perspective, and used to populate a 2-D histogram representing the shape features of the sketch image. The boundary between the support regions is calculated accordingly to each sketch point, which makes the approach scale invariant. We report results of objective evaluation of the proposed approach regarding robustness against noise, and comparative evaluation with three state-of-the-art methods, using an image database of scanned handwritten alphabets, digits, mathematical symbols and expressions. Experimental results show that the proposed method has competitive distinctiveness and robustness against noise.

A novel hand-drawn sketch descriptor based on the fusion of multiple features

“…The most similar work to ours can be found in paper [16] and [9], which have evolved from shape context [21]. But our work is different from these two approaches in two aspects: (1) the proposed approach uses the end points, branch points of the skeleton, and vertices of minimum enclosing rectangle of sketch as feature points.…”

“…The performance of the proposed descriptor (FMFD) and three state-of-the-art descriptors (shape context [21], CPDH [16][17] and Support Regions Descriptor (SRD) [9]) are evaluated and compared on several public sketch datasets. These three sketch descriptors are selected due to their popularity and similarity to the proposed approach.…”

“…Many shape description methods have been presented in the past decades [2]. In general, sketch descriptors can be classified into four categories: contour-based descriptor [3][4][5][6], region-based descriptor [7][8][9], skeleton-based descriptor [10], and hybrid-based descriptor [11][12][13]. Yang Cao et al [4] described a visual word using a triple (x, y, Θ) of the position (x, y) of an edge pixel ( edgel ) and the edgel orientationΘ at that position.…”

“…However, this representation inevitably results in the loss of position-invariant feature. Houssem Chatbri et al [9] introduced a descriptor which combines feature information in near and far support regions defined for each sketch point. Mathias Eitz et al [14][15] developed a bag-of-features sketch representation and represented sketches using local feature vectors that encode distributions of line orientation.…”

An Improved Thinning Algorithm for Fingerprint Recognition

“…Based on the literature review thinning algorithms either directly wok on grayscale image without binarization [12][13][14][15][16] or maximum algorithms required binary image as an input images [8,[17][18][19][20][21][22][23][24][25]. Three types of algorithm are worked for binary images: 1) sequential algorithms, 2) parallel algorithms, and 3) medial axis transform algorithms [8].…”