Blind distortion classification using content and perception based features

Praneeth, D; Venkatanath, N; Bh, Maruthi Chandrasekhar; Channappayya, Sumohana S.; Medasani, S.

doi:10.1109/qomex.2014.6982298

Cited by 1 publication

(4 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On the other hand, the proposed method provided very stable results for all distortion types considered. [20] 85.53 89.61 92.68 Mateusz [22] 82.42 90.34 84.55 Kang [32] 86.05 84.82 88.17 Kede [34] 89.67 85.63 90.28 Golestaneh [36] Finally, from Table 6, we can notice a significant decrease in performance of all methods, except DIIVINE, compared to their performance on the other datasets. Because, LIVEMD dataset contains images with multidistortion types making them difficult to recognize.…”

Section: Accepted Manuscript / Final Versionmentioning

confidence: 90%

“…The performance of our model is evaluated and compared to ten state-of-the-art methods, including distortion identification-based image verity and integrity evaluation (DIIVINE) [27], blind/referenceless image spatial quality evaluator (BRISQUE) [54], distortionspecific IQMs, COntent & Perception based features for DIstortion Classification (COPDIC) [20], Mateusz's method [22], Kang's method [32], Kede's method [34], Golestaneh's method [36], mask gated convolutional network (MGCN) [35] and Bianco's method [23]. The first four methods are based on hand-crafted features that are fed into a SVM classifier.…”

Section: Comparison and Discussionmentioning

confidence: 99%

“…Praneeth et al [20] proposed content and perceptionbased features for efficient distortion classification, called as COPDIC. Given an input image, features are derived from local block level characteristics to classify common distortion types using local mean removal, divisive normalization followed by natural scene statistics (NSS) features extraction.…”

Section: Distortion Identification As a Main Taskmentioning

confidence: 99%

“…It is therefore very interesting to develop an automatic method that mimic this HVS's capability. In recent years, much efforts have been devoted to the problem of identifying distortion in images and videos [18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34]. Most studies have focused on identifying a single type of distortion [26], however, images and videos can suffer from multiple distortions in most real-world applications [38], which is more challenging to address given the complex interactions and masking effect among distortions.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Deep multi-task learning for image/video distortions identification

Ameur

Fezza

Hamidouche

2021

Neural Comput & Applic

View full text Add to dashboard Cite

Identifying distortions in images and videos is important and useful in various visual applications, such as image quality enhancement and assessment techniques. Instead of applying them blindly, these techniques can be applied or adjusted depending on the type of distortion identified. In this paper, we propose a deep multi-task learning (MTL) model for identifying the types of distortion in both images and videos, considering both single and multiple distortions. The proposed MTL model is composed of one convolutional neural network (CNN) shared between all tasks and N parallel classifiers, where each classifier is dedicated to identify a type of distortion. The proposed architecture also allows to adjust the number of tasks according to the number of distortion types considered, making the solution scalable. The proposed method has been evaluated on natural scene images and laparoscopic videos databases, each presenting a rich set of distortions. The experimental results demonstrate that our model achieves the best performance among the state-of-art methods for both single and multiple distortions 1 .

show abstract