A spatial filtering approach to texture analysis

Coggins, James M.; Jain, Anil K.

doi:10.1016/0167-8655(85)90053-4

Cited by 199 publications

(76 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In these cases it outperforms the two best conventional feature sets. It can also be seen that the performance of all feature sets, excepting RING & WEDGE [11], are enhanced when multi-resolution data are used. In addition, the Varma and Zisserman texton [54] versions of the shape context [4] and chain code histogram [32] termed VZ-SC and VZ-CCH here; perform better than the two local texton feature sets (SCTH and CCHTH).…”

Section: Experimental Results 1) Pair-of-pairs Based Evaluation Exmentioning

confidence: 80%

“…The two best performing feature sets at these two resolutions, as reported in [17], i.e. Ring and Wedge Filters (RING & WEDGE) [11] and Multi-resolution Simultaneous Autoregressive Model (MRSAR) [37], were used as baselines for our comparison. These results are therefore shown separately in Fig.…”

Section: Experimental Results 1) Pair-of-pairs Based Evaluation Exmentioning

confidence: 99%

“…Inspired Bold fonts indicate the highest performance across the three edge detectors. [37] R&W [11] CCHTH VZ-CCH SCTH VZ-SC PMIF AR (%) [17] Bold fonts indicate the highest performance across different feature sets.…”

Section: ) Incorporating Other Image Characteristicsmentioning

confidence: 99%

“…The first comprises the type of 2nd-order statistics encoded in power spectra. They are often used by filter-based features [11], [17], [19] and encode both long-range and short-range periodicities. The second category concerns the HOS available from local image patches, i.e.…”

Section: The Importance Of Three Types Of Data To Texture Perceptionmentioning

confidence: 99%

“…In this experiment, the five best feature sets investigated in [19], namely, VZ-NBRHD [54], MRSAR [37], LBPBASIC [40], LBPHF [1] and RING & WEDGE [11], were utilised as baselines. The 7 measures obtained using the feature sets are shown in Fig.…”

Section: ) Retrieval Based Evaluation Experimentsmentioning

confidence: 99%

See 4 more Smart Citations

Perceptually Motivated Image Features Using Contours

Dong

Chantler

2016

IEEE Trans. on Image Process.

View full text Add to dashboard Cite

Abstract-Dong et al. examined the ability of 51 computational feature sets to estimate human perceptual texture similarity, however, none performed well for this task. While it is well-known that the human visual system is extremely adept at exploiting longer-range aperiodic (and periodic) "contour" characteristics in images, none of the investigated feature sets exploit higher order statistics (HOS) over larger image regions (>19×19 pixels). We therefore hypothesise that long-range HOS, in the form of contour data, are useful for perceptual texture similarity estimation.We present the results of a psychophysical experiment that shows that contour data are more important, than local image patches, or global 2nd-order data, to human observers for this task.Inspired by this finding, we propose a set of perceptually motivated image features (PMIF) that encode the long-range HOS computed from spatial and angular distributions of contour segments. We use two perceptual texture similarity estimation tasks to compare PMIF against the 51 feature sets referred to above and four commonly used contour representations. This new feature set is also examined in the context of two additional tasks: sketch-based image retrieval and natural scene recognition. The results show that the proposed feature set performs better, or at least comparably to, all the other feature sets. We attribute this promising performance to the fact that the proposed feature set exploits both short-range and long-range HOS.

show abstract