Indoor versus Outdoor Scene Classification Using Probabilistic Neural Network

Abstract

We propose a method for indoor versus outdoor scene classification using a probabilistic neural network (PNN). The scene is initially segmented (unsupervised) using fuzzy C-means clustering (FCM) and features based on color, texture, and shape are extracted from each of the image segments. The image is thus represented by a feature set, with a separate feature vector for each image segment. As the number of segments differs from one scene to another, the feature set representation of the scene is of varying dimension. Therefore a modified PNN is used for classifying the variable dimension feature sets. The proposed technique is evaluated on two databases: IITM-SCID2 (scene classification image database) and that used by Payne and Singh in 2005. The performance of different feature combinations is compared using the modified PNN.

Efficient Circular Thresholding

“…Given an input image, the purpose is to automatically identify whether the picture is taken indoors or outdoors. Gupta et al [24] used a probabilistic neural network classifier applied to a variety of colour, texture and shape image features. They also tested Payne and Singh's method [25] in which a multiresolution estimate of edge straightness is made.…”

“…For the application of indoor/outdoor image classification, we use the IITM-SCID2 scene classification image database [24] which has been constructed to discriminate between indoor and outdoor scenes. It contains 193/200 indoor/outdoor training images and 249/260 indoor/outdoor test images.…”

Exploiting Depth Information for Indoor-Outdoor Scene Classification

“…Most of the works on indoor-outdoor scene classification rely on low-level features based on colour, like histograms [7,17,18,20,21], and moments [10,16]. Many approaches are also based on textures [2,10,16,17,18,21,22] and/or on edges [7,14,16,21].…”

“…Many approaches are also based on textures [2,10,16,17,18,21,22] and/or on edges [7,14,16,21]. In a few works more complex features are also used, for example based on entropy of the pixel values [21], or shape [10].…”

“…Many approaches are also based on textures [2,10,16,17,18,21,22] and/or on edges [7,14,16,21]. In a few works more complex features are also used, for example based on entropy of the pixel values [21], or shape [10]. Features are extracted either from the whole image [21], or from regions obtained by a predefined rectangular block subdivision [2,14,17,18,20] or by image segmentation [7,10].…”