Improving Color Constancy Using Indoor–Outdoor Image Classification

Abstract

In this work, we investigate how illuminant estimation techniques can be improved, taking into account automatically extracted information about the content of the images. We considered indoor/outdoor classification because the images of these classes present different content and are usually taken under different illumination conditions. We have designed different strategies for the selection and the tuning of the most appropriate algorithm (or combination of algorithms) for each class. We also considered the adoption of an uncertainty class which corresponds to the images where the indoor/outdoor classifier is not confident enough. The illuminant estimation algorithms considered here are derived from the framework recently proposed by Van de Weijer and Gevers. We present a procedure to automatically tune the algorithms' parameters. We have tested the proposed strategies on a suitable subset of the widely used Funt and Ciurea dataset. Experimental results clearly demonstrate that classification based strategies outperform general purpose algorithms.

Scene classification in compressed and constrained domain

“…In this section we show preliminary results obtained by applying the proposed approach on different opposite classes of scene at superordinate level of description: open against closed, indoor against outdoor. These classes may be useful to properly address some parameters of IGP employed within imaging devices [1,6]. Moreover, we present a simple extension of the proposed method to work with multiple classes by just considering three classes usually managed in some way by a digital camera or a mobile phone: landscape, document and portraits.…”

“…Scenes belonging to these three classes are usually acquired by a consumer imaging device; the class inferred through a scene recognition engine may be useful for different tasks within IGP (e.g. colour constancy [6], optimised compression [9], etc. ).…”

Color constancy using 3D scene geometry

“…To cope with this problem, higher level visual information is taken into account recently [1,9,20]. In [20], the image is modeled as a mixture of semantic classes, such as sky, grass, road and buildings.…”

“…Illuminant estimation is steered by different classes by evaluating the likelihood of the semantic content. Similarly, indoor-outdoor image information is used in [1]. Alternatively, image statistics are used in [9] to improve color constancy.…”

“…Hence, this indicates that there is a correlation between image statistics, scene types and color constancy algorithms. Knowing the corresponding class of an image a priori leads to improved color constancy algorithms [1,9,20]. However, the more general case of unconstrained color constancy can not be solved if the classification of the input images is not directly incorporated in the algorithm.…”