Reflectance transformation imaging (RTI) is one of the most widely used techniques in order to digitize and analyze material appearance of a surface, finding a great level of utility and applicability in the field of cultural heritage as well as in industrial applications. To date, most of the methods used to process (model and relight) RTI data assume only one light direction for all pixels as well as a single light source-surface distance for the entire image, following the model of a very distant (far) light source. This assumption does not hold in practice. Indeed, the light sources commonly used in RTI acquisitions (spotlight/photo flash) induce to a non-uniform illumination of the surface. This is caused by the variation of incidence angles and per-point distances which directly affect the amount of light received by the surface. We propose a novel pixel-wise methodology for improving lighting based on illumination laws that allows one to correct both loss of energy due to the distance variation as well as the elevation angle. We show the efficiency of the proposed method on RTI acquisitions performed on cultural heritage objects and a manufactured surface. We show that our method corrects the effects of non-uniform illumination and leads to improve the relighting commonly associated with RTI.

Abstract In this paper, we propose a methodology for the fusion of shape from focus and reflectance transformation imaging. This fusion of two seemingly disparate methods of computational imaging is proposed with the purpose of leveraging their strengths in understanding overall surface structure (low-frequency detail) and surface texture/micro-geometry (high-frequency detail), respectively. This fusion is achieved by our new proposal of the integration of varying light images at different focus distances. We compare three methods of integration: the mean gradient response, the maximum gradient response, and the full vector gradient (FVG). The validation of the tested methods was conducted using different focus measure window sizes and multi-light integration methods to provide a clear demonstration of the effectiveness of the proposed method. The FVG is determined to provide a higher-quality shape recovery of a complex object with the trade-off of increasing the scope of the image acquisition.

Texture characterization from the metrological point of view is addressed in order to establish a physically relevant and directly interpretable feature. In this regard, a generic formulation is proposed to simultaneously capture the spectral and spatial complexity in hyperspectral images. The feature, named relative spectral difference occurrence matrix (RSDOM) is thus constructed in a multireference, multidirectional, and multiscale context. As validation, its performance is assessed in three versatile tasks. In texture classification on HyTexiLa, content-based image retrieval (CBIR) on ICONES-HSI, and land cover classification on Salinas, RSDOM registers 98.5% accuracy, 80.3% precision (for the top 10 retrieved images), and 96.0% accuracy (after post-processing) respectively, outcompeting GLCM, Gabor filter, LBP, SVM, CCF, CNN, and GCN. Analysis shows the advantage of RSDOM in terms of feature size (a mere 126, 30, and 20 scalars using GMM in order of the three tasks) as well as metrological validity in texture representation regardless of the spectral range, resolution, and number of bands.

Gradient extraction is important for a lot of metrological applications such as Control Quality by Vision. In this work, we propose a full-vector gradient for multi-spectral sensors. The full-vector gradient extends Di Zenzo expression to take into account the non-orthogonality of the acquisition channels thanks to a Gram matrix. This expression is generic and independent from channel count. Results are provided for a color and a multi-spectral snapshot sensor. Then, we show the accuracy improvement of the gradient calculation by creating a dedicated objective test and from real images.

A great deal of features detectors and descriptors are proposed every years for several computer vision applications. In this paper, we concentrate on dense detector applied to different descriptors. Eight descriptors are compared, three from gradient based family (SIFT, SURF, DAISY), others from binary category (BRIEF, ORB, BRISK, FREAK and LATCH). These descriptors are created and defined with certain invariance properties. We want to verify their invariances with various geometric and photometric transformations, varying one at a time. Deformations are computed from an original image. Descriptors are tested on five transformations: scale, rotation, viewpoint, illumination plus reflection. Overall, descriptors display the right invariances. This paper's objective is to establish a reproducible protocol to test descriptors invariances.