Performance evaluation of local descriptors for maximally stable extremal regions

Lee, Man Hee; Park, In Kyu

doi:10.1016/j.jvcir.2017.05.008

Cited by 16 publications

(10 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Specifically, maximally stable extremal region (MSER) detector is used to sample the interest points. MSER is an interest region based detector which has proven to be effective among its variant as it yields the best score in terms of effectiveness and efficiency in recent study [40]. MSER may provide a more discriminative interest points on food categories that have very strong mixture of ingredients as it may represent the irregular shape of foods, typically in parallelograms.…”

Section: Stage 1: Interest Points Detectionmentioning

confidence: 99%

“…MSER detector however was not built with its own descriptor. The empirical study conducted in [40] to evaluate the descriptors for MSER have suggested that the speeded-up robust feature (SURF) descriptor is very close to real-time applications. This is because SURF used integral image and Haar wavelet to approximate gradient information and very minimum of noises were generated.…”

Section: Stage 2: Feature Descriptionsmentioning

confidence: 99%

See 1 more Smart Citation

Fuzzy encoding with hybrid pooling for visual dictionary in food recognition

Razali

Manshor

Halin

et al. 2021

IJEECS

View full text Add to dashboard Cite

<span>Tremendous number of f food images in the social media services can be exploited by using food recognition for healthcare benefits and food industry marketing. The main challenges in food recognition are the large variability of food appearance that often generates a highly diverse and ambiguous descriptions of local feature. Ironically, the ambiguous descriptions of local feature have triggered information loss in visual dictionary constructions from the hard assignment practices. The current method based on hard assignment and Fisher vector approach to construct visual dictionary have unexpectedly cause errors from the uncertainty problem during visual word assignation. This research proposes a method of combination in soft assignment technique by using fuzzy encoding approach and maximum pooling technique to aggregate the features to produce a highly discriminative and robust visual dictionary across various local features and machine learning classifiers. The local features by using MSER detector with SURF descriptor was encoded by using fuzzy encoding approach. Support vector machine (SVM) with linear kernel was employed to evaluate the effect of fuzzy encoding. The results of the experiments have demonstrated a noteworthy classification performance of fuzzy encoding approach compared to the traditional approach based on hard assignment and Fisher vector technique. The effects of uncertainty and plausibility were minimized along with more discriminative and compact visual dictionary representation.</span>

show abstract

Section: Stage 1: Interest Points Detectionmentioning

confidence: 99%

Section: Stage 2: Feature Descriptionsmentioning

confidence: 99%

Fuzzy encoding with hybrid pooling for visual dictionary in food recognition

Razali

Manshor

Halin

et al. 2021

IJEECS

View full text Add to dashboard Cite

show abstract

“…Then, all the new set of the quantity of ER will be updated in cell array before the features are extracted by using Speeded Up Robust Feature (SURF) descriptor. SURF is chosen to be paired with MSER due its balanced performance between accuracy and efficiency, less sensitive to noise and more practical for real time application [10], [11], [14]. Furthermore, SURF generates shorter length of feature vector that was reasoned for a speedy feature encoding process, produced a distinctive feature and robust to the geometric and photometric deformation.…”

Section: Flowchart Of Amser Algorithmmentioning

confidence: 99%

“…There are numerous types of local features in the literature. A research conducted by [10] employed the interest region detector by using Maximally Stable Extremal Region (MSER) to detect food interest points, considering MSER as among the best interest region detector in term of effectiveness and efficiency [11]. MSER detects a set of connected regions from an image to define the extremal regions (ER).…”

Section: Introductionmentioning

confidence: 99%

“…MSER encountered the similar problem as the other interest points detector. In fact, according to [11], [14], MSER detect even smaller number of interest points among the interest points detector. Therefore, this study proposed an adaptive approach for ER detection in MSER (aMSER) that choose appropriate MSER parameter configuration to increase the density of ER on the texture-less food images.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Improving Invisible Food Texture Detection by using Adaptive Extremal Region Detector in Food Recognition

Razali¹

2019

IJATCSE

View full text Add to dashboard Cite

The advancement of mobile technology with reasonable cost has indulge the mobile phone users to photograph foods and shared in social media. Since that, food recognition has become emerging research area in image processing and machine learning. Food recognition provides an automatic identification of the types of foods from an image. Then, further analysis in food recognition is performed to approximate the calories and nutritional information that can be used for health-care purposes. The interest region-based detector by using Maximally Stable Extremal Region (MSER) may provides distinctive interest points by representing the arbitrary shape of foods through global segmentation especially the food images with strong mixture of ingredients. However, the classification performance on food categories with less diverse texture food images by using MSER are obviously low compared to the other food categories that have more noticeable texture. The texture-less food objects were suffered from small number of extremal regions (ER) detection beside having low image brightness and small resolutions. Therefore, this paper proposed an adaptive interest regions detection by using MSER (aMSER) that provide a mechanism to choose appropriate MSER parameter configuration to increase the density of interest points on the targeted food images. The features are described by using Speeded-up Robust Feature Transform (SURF) and encoded by using Bag of Features (BoF) model. The classification is performed by using Linear Support Vector Machine and yield 84.20% classification rate on UEC100-Food dataset with competitive number of ER and computation cost.

show abstract