Michela Lecca

Energy-driven path search for Termite Retinex

The human color sensation depends on the local and global spatial arrangements of the colors in the scene. Emulating this dependence requires the exploration of the image in search of a white reference. The algorithm Termite Retinex explores the image by a set of paths resembling traces of a swarm of termites. Starting from this approach, we develop a novel spatial exploration scheme where the termite paths are local minimums of an energy function, which depend on the image visual content. The energy is designed to favor the visitation of regions containing information relevant to the color sensation while minimizing the coverage of less essential regions. This exploration method contributes to the investigation of the spatial properties of the color sensation and, to the best of our knowledge, is the first model relying on mathematical global conditions for the Retinex paths. The experiments show that the estimation of the color sensation obtained by means of the proposed spatial sampling is a valid alternative to the one based on Termite Retinex.

A population-based approach to point-sampling spatial color algorithms

Inspired by the behavior of the human visual system, spatial color algorithms perform image enhancement by correcting the pixel channel lightness based on the spatial distribution of the intensities in the surrounding area. The two visual contrast enhancement algorithms RSR and STRESS belong to this family of models: they rescale the input based on local reference values, which are determined by exploring the image by means of random point samples, called sprays. Due to the use of sampling, they may yield a noisy output. In this paper, we introduce a probabilistic formulation of the two models: our algorithms (RSR-P and STRESS-P) rely implicitly on the whole population of possible sprays. For processing larger images, we also provide two approximated algorithms that exploit a suitable target-dependent space quantization. Those spray population-based formulations outperform RSR and STRESS in terms of the processing time required for the production of noiseless outputs. We argue that this population-based approach, which can be extended to other members of the family, complements the sampling-based approach, in that it offers not only a better control in the design of approximated algorithms, but also additional insight into individual models and their relationships. We illustrate the latter point by providing a model of halo artifact formation.

An Event-Driven Ultra-Low-Power Smart Visual Sensor

In this paper, we present an ultra-low-power smart visual sensor architecture. A 10.6-μW low-resolution contrast-based imager featuring internal analog preprocessing is coupled with an energy-efficient quad-core cluster processor that exploits near-threshold computing within a few milliwatt power envelope. We demonstrate the capability of the smart camera on a moving object detection framework. The computational load is distributed among mixed-signal pixel and digital parallel processing. Such local processing reduces the amount of digital data to be sent out of the node by 91%. Exploiting context aware analog circuits, the imager only dispatches meaningful postprocessed data to the processing unit, lowering the sensor-to-processor bandwidth by 31× with respect to transmitting a full pixel frame. To extract high-level features, an event-driven approach is applied to the sensor data and optimized for parallel runtime execution. A 57.7× system energy saving is reached through the event-driven approach with respect to frame-based processing, on a low-power MCU node. The near-threshold parallel processor further reduces the processing energy cost by 6.64×, achieving an overall system energy cost of 1.79 μJ per frame, which results to be 21.8× and up to 383× lower than, respectively, an event-based imaging system based on an asynchronous visual sensor and a traditional frame-based smart visual sensor.

A New Region-based Active Contour Model for Object Segmentation

We present a novel region-based active contour model that segments one or more image regions that are visually similar to an object of interest, said prior. The region evolution equation of our model is defined by a simple heuristic rule and it is not derived by minimizing an energy functional, as in the classic variational approaches. The prior and the evolving region are described by the probability density function (pdf) of a photometric feature, as color or intensity. The heuristic rule deforms an initial region of the image in order to equalize pointwise the pdfs of the prior and of the region. Such heuristic rule can be modeled by many mathematical monotonic decreasing functions, each defining an evolution equation for the initial image region. The choice of a particular function is remitted to the user, that in this way can even integrate a priori knowledge possibly useful to break down the computational charge of the method and to increase the detection accuracy. Here we propose two different evolution equations for the general purpose of prior detection without a priori information and we discuss empirically the performances of our model on real-world and synthetic datasets. These experiments show that our model is a valid alternative to the classic models.

On edge-aware path-based color spatial sampling for Retinex: from Termite Retinex to Light Energy-driven Termite Retinex

Abstract. Retinex theory estimates the human color sensation at any observed point by correcting its color based on the spatial arrangement of the colors in proximate regions. We revise two recent path-based, edge-aware Retinex implementations: Termite Retinex (TR) and Energy-driven Termite Retinex (ETR). As the original Retinex implementation, TR and ETR scan the neighborhood of any image pixel by paths and rescale their chromatic intensities by intensity levels computed by reworking the colors of the pixels on the paths. Our interest in TR and ETR is due to their unique, content-based scanning scheme, which uses the image edges to define the paths and exploits a swarm intelligence model for guiding the spatial exploration of the image. The exploration scheme of ETR has been showed to be particularly effective: its paths are local minima of an energy functional, designed to favor the sampling of image pixels highly relevant to color sensation. Nevertheless, since its computational complexity makes ETR poorly practicable, here we present a light version of it, named Light Energy-driven TR, and obtained from ETR by implementing a modified, optimized minimization procedure and by exploiting parallel computing.

A 30 µW 30 fps 110 × 110 Pixels Vision Sensor Embedding Local Binary Patterns

We present a 110 × 110 pixel vision sensor that computes the Local Binary Patterns (LBPs) of an imaged scene with a power consumption of 30 μW at 30 fps. The LBP of a given pixel is a binary vector, encoding the direction and sign of image contrast with respect to its neighbors. Each LBP provides a visual description of an images local structure that is widely used for texture and object recognition. In the sensor proposed here, each pixel detects its corresponding LBP with respect to its four neighboring pixels and saves this information into a digital map using 6 bits to encode each pixel. The operation is executed during the exposure time and requires 83 pW/pixel · frame to be computed. The chip is implemented in a 0.35 μm CMOS featuring 34 T square pixels with 26 μm pitch. We illustrate some examples of image description based on the LBPs output by the sensor.

Illuminant Change Estimation via Minimization of Color Histogram Divergence

We present a new method for computing the change of light possibly occurring between two pictures of the same scene. We approximate the illuminant variation with the von Kries diagonal transform and estimate it by minimizing a functional that measures the divergence between the image color histograms. Our approach shows good performances in terms of accuracy of the illuminant change estimation and of robustness to pixel saturation and Gaussian noise. Moreover we illustrate how the method can be applied to solve the problem of illuminant invariant image recognition.

GRASS: A Gradient-Based Random Sampling Scheme for Milano Retinex

Retinex is an early and famous theory attempting to estimate the human color sensation derived from an observed scene. When applied to a digital image, the original implementation of retinex estimates the color sensation by modifying the pixels channel intensities with respect to a local reference white, selected from a set of random paths. The spatial search of the local reference white influences the final estimation. The recent algorithm energy-driven termite retinex (ETR), as well as its predecessor termite retinex, has introduced a new path-based image aware sampling scheme, where the paths depend on local visual properties of the input image. Precisely, the ETR paths transit over pixels with high gradient magnitude that have been proved to be important for the formation of color sensation. Such a sampling method enables the visit of image portions effectively relevant to the estimation of the color sensation, while it reduces the analysis of pixels with less essential and/or redundant data, i.e., the flat image regions. While the ETR sampling scheme is very efficacious in detecting image pixels salient for the color sensation, its computational complexity can be a limit. In this paper, we present a novel Gradient-based RAndom Sampling Scheme that inherits from ETR the image aware sampling principles, but has a lower computational complexity, while similar performance. Moreover, the new sampling scheme can be interpreted both as a path-based scanning and a 2D sampling.

Energy-efficient design of an always-on smart visual trigger

In this work, we present the design of an always-on smart visual trigger. To maximize the energy-efficiency, the whole system is kept in stand-by mode until a significant information is detected by the early-processing of the low-power imager. Within two considered scenarios of vehicle detection, the system runs at minimal power consumption for 84% and 39% of the time. When active, the generation of triggers due to relevant events is conducted by analyzing the trajectory of multiple tracked objects. A parallel event-driven implementation speeds-up the digital computation and leads to a duty cycle below 1% over the frame period. The optimized power management is enabled by defining an always-on camera interface for the System-on-Chip (SoC) processor, which is able to individually activate both the sensor and the processor while running at minimal power consumption. In the considered case-study of vehicle detection, an estimated power consumption of up to 23μW is accounted, depending on the context-activity, and the smart triggers fails one detection over 72 moving vehicles.

Ultra-low power high-dynamic range color pixel embedding RGB to r-g chromaticity transformation

This work describes a novel color pixel topology that converts the three chromatic components from the standard RGB space into the normalized r-g chromaticity space. This conversion is implemented with high-dynamic range and with no dc power consumption, and the auto-exposure capability of the sensor ensures to capture a high quality chromatic signal, even in presence of very bright illuminants or in the darkness. The pixel is intended to become the basic building block of a CMOS color vision sensor, targeted to ultra-low power applications for mobile devices, such as human machine interfaces, gesture recognition, face detection. The experiments show that significant improvements of the proposed pixel with respect to standard cameras in terms of energy saving and accuracy on data acquisition. An application to skin color-based description is presented.