Raymond Phan

Content-based retrieval of logo and trademarks in unconstrained color image databases using Color Edge Gradient Co-occurrence Histograms

In this paper, we present an algorithm that extends the Color Edge Co-occurrence Histogram (CECH) object detection scheme on compound color objects, for the retrieval of logos and trademarks in unconstrained color image databases. We introduce more accurate information to the CECH, by virtue of incorporating color edge detection using vector order statistics. This produces a more accurate representation of edges in color images, as compared to the simple color pixel difference classification of edges seen with the CECH. Our proposed method is thus reliant on edge gradient information, and so we call it the Color Edge Gradient Co-occurrence Histogram (CEGCH). We also introduce a color quantization method based in the hue-saturation-value (HSV) color space, illustrating that it is a more suitable scheme of quantization for image retrieval, compared to the color quantization scheme introduced with the CECH. Experimental results demonstrate that the CEGCH and the HSV color quantization scheme is insensitive to scaling, rotation, and partial deformations, and outperforms the use of the CECH in image retrieval, with higher precision and recall. We also perform experiments on a closely related algorithm based on the Color Co-occurrence Histogram (CCH) and demonstrate that our algorithm is also superior in comparison, with higher precision and recall.

Robust Semi-Automatic Depth Map Generation in Unconstrained Images and Video Sequences for 2D to Stereoscopic 3D Conversion

We describe a system for robustly estimating synthetic depth maps in unconstrained images and videos, for semi-automatic conversion into stereoscopic 3D. Currently, this process is automatic or done manually by rotoscopers. Automatic is the least labor intensive, but makes user intervention or error correction difficult. Manual is the most accurate, but time consuming and costly. Noting the merits of both, a semi-automatic method blends them together, allowing for faster and accurate conversion. This requires user-defined strokes on the image, or over several keyframes for video, corresponding to a rough estimate of the depths. After, the rest of the depths are determined, creating depth maps to generate stereoscopic 3D content, with Depth Image Based Rendering to generate the artificial views. Depth map estimation can be considered as a multi-label segmentation problem: each class is a depth. For video, we allow the user to label only the first frame, and we propagate the strokes using computer vision techniques. We combine the merits of two well-respected segmentation algorithms: Graph Cuts and Random Walks. The diffusion from Random Walks, with the edge preserving of Graph Cuts should give good results. We generate good quality content, more suitable for perception, compared to a similar framework.

Semi-automatic 2D to 3D image conversion using scale-space Random Walks and a graph cuts based depth prior

In this paper, we present a semi-automated method for converting conventional 2D images into stereoscopic 3D. User-defined strokes corresponding to a rough estimate of the depth values in the scene are defined for the image of interest. With these, our system determines the depth values for the rest of the image, producing a depth map that can be used to create stereoscopic 3D image pairs. Our work is based on a similar scheme, using the Random Walks segmentation paradigm. However, the related work is quite complex, with many processing steps required to produce the final stereoscopic image pair. Combined with its evident shortcomings, but noting the merits, we propose a system employing Random Walks, while incorporating information from the popular Graph Cuts segmentation paradigm. Thus, a final cohesive depth map is produced, combining the merits of both. The results show that we can produce good quality stereoscopic image pairs, while using a much more simplified method in comparison to the related work.

Semi-automatic synthetic depth map generation for video using random walks

We present a method for easily generating depth maps from monoscopic (i.e. “2D”) video footage in order to convert them into stereoscopic, or “3D”, footage. Our method uses user-defined strokes for a number of keyframes in the original footage and interpolates between the keyframes to provide a sparse labelling for each frame. We then apply the Random Walks algorithm to the footage to provide depth estimates based on the input provided by the user. These depth maps can then be used to generate novel views through depth-based image rendering.

Semi-automatic 2D to 3D image conversion using a hybrid Random Walks and graph cuts based approach

In this paper, we present a semi-automated method for converting conventional 2D images to stereoscopic 3D. User-defined strokes that correspond to a rough estimate of the depth values in the scene are defined for the image of interest. With these strokes, our system thus determines what the depth values are for the rest of the image, producing a depth map that is ultimately used to create a stereoscopic image pair. Our work is based on a similar scheme which employs Random Walks. However, the related work is quite complex, with many processing steps required to produce the final stereoscopic image pair. Combined with the evident shortcomings of the related work, but noting the merits of Random Walks, we propose a system that is a hybrid between Random Walks, and the popular Graph Cuts segmentation paradigm. Both segmentation algorithms are used to generate a final cohesive depth map, thus combining the merits of both frameworks together. The generated results show that we can produce good quality stereoscopic image pairs, while using a much more simplified method in comparison to the related Random Walks scheme.

Unconstrained logo and trademark retrieval in general color image databases using Color Edge Gradient Co-occurrence Histograms

In this paper, we present a logo and trademark retrieval system for general, unconstrained, color image databases, extending the color edge co-occurrence histogram (CECH) object detection scheme. We introduce more accurate information to the CECH, by virtue of incorporating color edge detection using vector order statistics. This produces a more accurate representation of edges in color images, in comparison to the simple color pixel difference classification of edges as seen in the CECH. Our proposed method is thus reliant on edge gradient information, thus we call it the color edge gradient co-occurrence histogram (CEGCH). We also introduce a novel color quantization scheme based in the hue-saturation-value (HSV) color space, illustrating that it is more suitable for image retrieval in comparison to the color quantization scheme introduced with the CECH. Results illustrate that our retrieval system retrieves logos and trademarks with good accuracy, outperforming the use of the CECH in image retrieval with higher precision and recall.

Interactive video GrowCut: A semi-automated video object segmentation framework using cellular automata

This paper illustrates a simple, yet effective semi-automated object segmentation framework over video sequences. This is through an extension of the GrowCut framework, an image segmentation scheme based on cellular automata. We describe how GrowCut is extended to video sequences, as well as providing our own improvements and addressing problematic areas to the original formulation. This provides a good increase in accuracy and creates the main goal of this work. It will be shown that the original algorithm adapts quite nicely to video object segmentation, and can achieve very good results using both synthetic and real video footage, obtained from different sources.

Colour logo and trademark detection in unconstrained images using colour edge gradient co-occurrence histograms

In this paper, we present an extension of the Colour Edge Co-occurence Histogram (CECH) object detection scheme for detecting logos and trademarks in unconstrained colour images. We introduce more accurate information to the CECH by virtue of incorporating colour edge detection using vector order statistics, producing a more accurate representation of edges in images, as compared to the simple colour difference edge classification which is done in the CECH. Our proposed method is thus reliant on edge gradient information, and so we call it the Colour Edge Gradient Co-occurrence Histogram (CEGCH). We also illustrate a colour quantization scheme based in the Hue-Saturation-Value (HSV) colour space, illustrating that it is more suitable for logo and trademark detection in comparison to the colour quantization scheme used with the CECH. Results illustrate that the CEGCH detects logos and trademarks with greater accuracy in comparison to the CECH.

Edge-aware temporally consistent SimpleFlow: Optical flow without global optimization

Optical Flow is a very important topic in computer vision, with applications in object tracking, motion estimation and video compression. Recently, Tao et al. proposed the Simple-Flow algorithm - a non-iterative method whose running times increase sublinearly with the number of pixels. SimpleFlow does not use global optimization and uses only local evidence, achieving significant speedups in parallel programming environments. With this, we extend SimpleFlow by taking advantage of edge-aware filtering methods to increase accuracy, and allow SimpleFlow to be temporally consistent over video. The combination of temporal consistency and edge-aware filtering will inevitably create a smooth motion field across the video. We show results illustrating an increase in accuracy in comparison to the original SimpleFlow framework, for images and multi-frame datasets.

Adaptive 2D to 3D image conversion using a hybrid Graph Cuts and Random Walks approach

In this paper, we propose an adaptive method for 2D to 3D conversion of images using a user-aided process based on Graph Cuts and Random Walks. Given user-defined labeling that correspond to a rough estimate of depth, the system produces a depth map which, combined with a 2D image can be used to synthesize a stereoscopic image pair. The work presented here is an extension of work done previously combining the popular Graph Cuts and Random Walks image segmentation algorithms. Specifically, we have made the previous approach adaptive, as well as improved the quality of the results. This is achieved by feeding information from the Graph Cuts result into the Random Walks process at two different stages, and using edge and spatial information to adapt various weights. The results show that we can produce good quality stereoscopic 3D image pairs using a simple yet adaptive approach.