No abstract

No abstract

No abstract

Selfies have become commonplace. More and more people take pictures of themselves, and enjoy enhancing these pictures using a variety of image processing techniques. One specific functionality of interest is automatic skin and hair segmentation, as this allows for processing one's skin and hair separately. Traditional approaches require user input in the form of fully specified trimaps, or at least of "scribbles" indicating foreground and background areas, with high-quality masks then generated via matting. Manual input, however, can be difficult or tedious, especially on a smartphone's small screen. In this paper, we propose the use of fully convolutional networks (FCN) and fully-connected CRF to perform pixel-level semantic segmentation into skin, hair and background. The trimap thus generated is given as input to a standard matting algorithm, resulting in accurate skin and hair alpha masks. Our method achieves state-of-the-art performance on the LFW Parts dataset [1]. The effectiveness of our method is also demonstrated with a specific application case.

This paper presents a 4-GHz all-digital fractional-N PLL with a low-power TDC operating at low-rate retimed reference clocks, a compensator preventing big phase-error downfalls, and a loop settling monitor. Two retimed reference clocks, nCKR and pCKR, are employed in the TDC to estimate the fractional phase error between the low-rate reference and high-rate oscillator clocks. Applying the retimed reference clocks does not only reduce a dynamic power in its delay chain, but simplify a fractional phase-error correction. The phase-error compensator is introduced to avoid big phase-error downfalls caused by large output glitches originating from a high-speed accumulator. In addition, a loop-settling monitor is invented to allow the DCO operation mode to be shifted seamlessly and fast. By consuming 9.6 mW, the ADPLL achieves in-band phase noise, integrated noise, and 740 ns settling time.

No abstract

We present an algorithm that finds planar structures in a Manhattan world from two pictures taken from different viewpoints with unknown baseline. The Manhattan world assumption constrains the homographies induced by the visible planes on the image pair, thus enabling robust reconstruction. We extend the T-linkage algorithm for multistructure discovery to account for constrained homographies, and introduce algorithms for sample point selection and orientation-preserving cluster merging. Results are presented on three indoor data set, showing the benefit of the proposed constraints and algorithms.

This letter presents a CMOS RF front-end operating in a subthreshold region for low-power Band-III mobile TV applications. The performance and feasibility of the RF front-end are verified by integrating with a low-IF RF tuner fabricated in a 0.13-μm CMOS technology. The RF front-end achieves the measured noise figure of 4.4 dB and a wide gain control range of 68.7 dB with a maximum gain of 54.7 dB. The power consumption of the RF front-end is 13.8 mW from a 1.2 V supply.