In order to characterize the performance of visible digital imaging systems in the laboratory, in the field and in simulation, the use of fractal test-targets has been optimized. This work is based on the previous achievements in the use of binary fractal targets (2014) and the Corner-Point (CP) resolution criterion (2017), for DRI range modeling of optronic cameras. The principle is to resume from the process of multi-scale fractal calculation of the binary target, to extend it in the case of a multi-level of gray. The distribution of CP contrasts by scale is then adapted to two constraints, on the one hand the measurement accuracy and on the other hand the criterion definition of the operational task evaluated for the camera. A target will be specifically designed to accommodate an operational need, such as the identification of vehicles or handheld weapon. The exploitation of the fractal target degraded by the imager is carried out by the comparison of CP by scales with the original target, after an image registration phase, facilitated by an original Yin-Yang design of the target at its lowest CP scale. The main metric for assessing DRI range is the Resolved Contrast Function (RCF), obtained from the multi-scale CP Probability of Correct Resolution. In the first part of the paper, the principles of design and exploitation of the target are presented, applied to an example of a DRI range assessment of a camera coupled to image restoration processing. In a second part, the use of this evaluation technique is developed in the example of digital image fusion systems, from two bands with its own optronic characteristics and some non-linear digital processing. This work, carried out in simulation using the FUSIM software, allows to establish a selection of the optimal combinations (pre-processing, fusion processing) offering the best RCF.
Range performance modeling of optronics imagers attempts to characterize the ability to resolve details in the image. Today, digital image processing is systematically used in conjunction with the optoelectronic system to correct its defects or to exploit tiny detection signals to increase performance. In order to characterize these processing having adaptive and non-linear properties, it becomes necessary to stimulate the imagers with test patterns whose properties are similar to the actual scene image ones, in terms of dynamic range, contours, texture and singular points. This paper presents an approach based on a Corner-Point (CP) resolution criterion, derived from the Probability of Correct Resolution (PCR) of binary fractal patterns. The fundamental principle lies in the respectful perception of the CP direction of one pixel minority value among the majority value of a 2×2 pixels block. The evaluation procedure considers the actual image as its multi-resolution CP transformation, taking the role of Ground Truth (GT). After a spatial registration between the degraded image and the original one, the degradation is statistically measured by comparing the GT with the degraded image CP transformation, in terms of localized PCR at the region of interest. The paper defines this CP criterion and presents the developed evaluation techniques, such as the measurement of the number of CP resolved on the target, the transformation CP and its inverse transform that make it possible to reconstruct an image of the perceived CPs. Then, this criterion is compared with the standard Johnson criterion, in the case of a linear blur and noise degradation. The evaluation of an imaging system integrating an image display and a visual perception is considered, by proposing an analysis scheme combining two methods: a CP measurement for the highly non-linear part (imaging) with real signature test target and conventional methods for the more linear part (displaying). The application to color imaging is proposed, with a discussion about the choice of the working color space depending on the type of image enhancement processing used.
The DGA (Delegation Generale de l'Armement) is interested in the determination of sky-ground characteristics. In particuliar, optical clouds properties as well as land-surface and sea-surface temperatures must be determined with accuracy. To obtain a statistical description of the cloud properties, we have created a cloud database called SALIC (SAtellite-LIdar-Clouds). Two algorithms, one for sea-surface temperature and one for land-surface temperature were recently included in the database. Three different kinds of measurements are used to built up the database: radiosoundings, ground-based lidar measurements, and satellite data obtained from the radiometer AVHRR3 boarded on the NOAA-16 polar orbiting satellite. This paper presents the results for a period covering two years.