Local Parallel Cross Pattern: A Color Texture Descriptor for Image Retrieval

Abstract

Riding the wave of visual sensor equipment (e.g., personal smartphones, home security cameras, vehicle cameras, and camcorders), image retrieval (IR) technology has received increasing attention due to its potential applications in e-commerce, visual surveillance, and intelligent traffic. However, determining how to design an effective feature descriptor has been proven to be the main bottleneck for retrieving a set of images of interest. In this paper, we first construct a six-layer color quantizer to extract a color map. Then, motivated by the human visual system, we design a local parallel cross pattern (LPCP) in which the local binary pattern (LBP) map is amalgamated with the color map in “parallel” and “cross” manners. Finally, to reduce the computational complexity and improve the robustness to image rotation, the LPCP is extended to the uniform local parallel cross pattern (ULPCP) and the rotation-invariant local parallel cross pattern (RILPCP), respectively. Extensive experiments are performed on eight benchmark datasets. The experimental results validate the effectiveness, efficiency, robustness, and computational complexity of the proposed descriptors against eight state-of-the-art color texture descriptors to produce an in-depth comparison. Additionally, compared with a series of Convolutional Neural Network (CNN)-based models, the proposed descriptors still achieve competitive results.

Local Ternary Cross Structure Pattern: A Color LBP Feature Extraction with Applications in CBIR

“…In addition, the open source implementation of the LBP feature map [9,10] can be downloaded from http://www.ee.oulu.fi/~{}gyzhao/. Referring to [6,22,23,27,40,49,50], all images in the dataset are considered as the query images to guarantee the accurateness and reproducibility. Referring to [6,22,23,27,40,49,50], the number of returned images on VisTex, KTH-2a, Face-95 and Corel-1000 is defined as 10, and the number of returned images on ETHZ and ZuBuD is defined as five because there are only five images in each class.…”

“…In addition, we provide the MSE values of 25 color quantization levels (A a* , A b* ) between the FLCQ and EICQ quantizers in the CIELAB color model on all six datasets in Appendix A. Table 5 reports the evaluations of the APR and ARR rates obtained by the proposed descriptors and a series of state-of-the-art color LBP descriptors including OCLBP [18], IOCLBP [19], maLBP [23] mdLBP [23], OC-LBP + CH [26], LPCP [27] on all six datasets. The best values are highlighted in bold.…”

Special Issue on Visual Sensors

“…Fareed et al [10] proposed a framework for salient region detection that uses appearance-based and regression-based schemes to reduce the computational complexity and focusing on the salient parts of the image. In this sense, Feng et al [11] proposed a texture descriptor for image retrieval designing a local parallel cross pattern in which the local binary pattern map is fused with the color map. In addition, Feng et al [12] proposed a hybrid histogram descriptor used for image retrieval.…”

Quantized Neighborhood Local Intensity Extrema Patterns for Image Retrieval

“…P. Banerjee et al extracted the mutual relationship among the pixels surrounding the central pixel (Banerjee et al, 2018). Authors proposed parallel cross pattern for retrieval purpose Intensity patterns are extracted for a 3x3 kernel (Feng et al, 2019). In this work, quantization is Cite The Article: L Koteswara Rao, P Rohini and P Sree Lakshmi (2020).…”