Serene Banerjee

In-Camera Automation of Photographic Composition Rules

At the time of image acquisition, professional photographers apply many rules of thumb to improve the composition of their photographs. This paper develops a joint optical-digital processing framework for automating composition rules during image acquisition for photographs with one main subject. Within the framework, we automate three photographic composition rules: repositioning the main subject, making the main subject more prominent, and making objects that merge with the main subject less prominent. The idea is to provide to the user alternate pictures obtained by applying photographic composition rules in addition to the original picture taken by the user. The proposed algorithms do not depend on prior knowledge of the indoor/outdoor setting or scene content. The proposed algorithms are also designed to be amenable to software implementation on fixed-point programmable digital signal processors available in digital still cameras.

Unsupervised automation of photographic composition rules in digital still cameras

When taking pictures, professional photographers apply photographic composition rules, e.g. rule of thirds. The rule of thirds says to place the main subjects center at one of four places: at 1/3 or 2/3 of the picture width from left edge, and 1/3 or 2/3 of the picture height from the top edge. This paper develops low-complexity unsupervised methods for digital still cameras to (1) segment the main subject and (2) realize the rule-of-thirds. The main subject segmentation method uses the auto-focus filter, opens the shutter aperture fully, and segments the resulting image. These camera settings place the main subject in focus and blur the rest of the image by diffused light. The segmentation utilizes the difference in frequency content between the main subject and blurred background. The segmentation does not depend on prior knowledge of the indoor/outdoor setting or scene content. The rule-of-thirds method moves the centroid of the main subject to the closest of the four rule-of-thirds locations. We first define an objective function that measures how close the main subject placement obeys the rule-of-thirds, and then reposition the main subject in order to optimize the objective function. For multiple main subjects, the proposed algorithm could be extended to use rule-of-triangles by adding an appropriate constraint.

Unsupervised merger detection and mitigation in still images using frequency and color content analysis

When taking pictures, professional photographers apply photographic composition rules, e.g. avoidance of mergers. A merger occurs when equally focused foreground and background regions appear to merge as one object. This paper presents an unsupervised algorithm that: (a) detects the main subject; (b) detects background objects merging with the main subject; and (c) reduces the visibility of merging background objects. Detection of the main subject requires automated adjustment of camera settings. The rest of the algorithm does not adjust or use the camera settings. The algorithm does not make assumptions about the scene setting (indoor/outdoor) or content. The algorithm is amenable to implementation on a fixed-point processor.

Tile-Based Transport of JPEG 2000 Images

Because it offers access to portions of the compressed data JPEG 2000 is not just a compression system but an alternative image representation. The JPEG 2000 image representation is better than uncompressed data in some cases and better than other compressed formats in most cases. The TRUEW system provides network access to portions of JPEG 2000 compressed files using tile-parts. This tile-part access is being standardized as part of JPEG 2000 Part 9.

A novel gradient induced main subject segmentation algorithm for digital still cameras

When taking pictures, professional photographers employ a variety of composition rules. In automating these rules, it is often first necessary to detect and segment the main subject. We propose an detection and segmentation algorithm that leverages the optics in a digital still camera. Based on where the user points the camera, an autofocus filter first puts the main subject in focus and takes a picture. Then, we open the shutter aperture to diffuse light from objects that are out-of-focus, which blurs the background, and take a second picture. Using the second picture, the resulting difference in the frequency content of the main subject and the background image is then used by the proposed algorithm to detect and segment the main subject. The algorithm does not depend on prior knowledge of the indoor/outdoor setting or scene content. Algorithm complexity is similar to that of a 5/spl times/5 filter.

Tuning JPEG2000 image compression for graphics regions

The new image compression standard, JPEG2000, provides higher compression rates for the same visual quality for grayscale and color images than JPEG. JPEG2000 is being adopted for image compression and transmission, eg, in mobile telephones, PDA, and wearable computers. Depending on the application, images may contain formatted text and graphics data. For graphics images at the same low bit rates, graphics compression methods, such as the graphics interchange format (GIF) and portable network graphics (PNG), outperform JPEG2000 in visual quality. In this paper, we describe problems associated with compressing graphics data with JPEG2000, and propose modifications to a JPEG2000 encoder to minimize distortion for color graphics data by using a model of the human visual system. The visual improvements are quantified by a new measure of distortion.

The transport of reversible and unreversible embedded wavelets (TRUEW)

Summary form only given. Transport of reversible and unreversible embedded wavelets (TRUEW) is a system for the portion delivery of compressed JPEG 2000 files over a network. A discussion on the development of the TRUEW architecture, the benefits of the TRUEW architecture, an efficiency analysis of tile-part and precint access to JPEG 2000 compressed images, and some comments on potential applications of interactive image access are presented. Computations show that for an example access pattern, TRUEW provides comparable performance in terms of disk accesses and data transferred with more complex precinct based systems.

VLIW DSP vs. superscalar implementation of a baseline 11.263 video encoder

A Very Long Instruction Word (VLIW) processor and a superscalar processor can execute multiple instructions simultaneously. A VLIW processor depends on the compiler and programmer to find the parallelism in the instructions, whereas a superscaler processor determines the parallelism at runtime. This paper compares TI TMS320C6700 VLIW digital signal processor (DSP) and SimpleScalar superscalar implementations of a baseline 11.263 video encoder in C. With level two C compiler optimization, a one-way issue superscalar processor is 7.5 times faster than the VLIW DSP for the same processor clock speed. The superscalar speedup from one-way to four-way issue is 2.88:1, and from four-way to 256-way issue is 2.43:1. To reduce the execution time on the C6700, we write assembly routines for sum-of-absolute-difference, interpolation, and reconstruction, and place frequently used code and data into on-chip memory. We use TIs discrete cosine transform assembly routines. The hand optimized VLIW DSP implementation is 61/spl times/ faster than the C version compiled with level two optimization. Most of the improvement was due to the efficient placement of data and programs in memory. The hand optimized VLIW implementation is 14% faster than a 256-way superscalar implementation without hand optimizations.