Image classification and cognition using contour curvature statistics

Although the higher order mechanisms behind object representation and classification in the visual system are still not well understood, there are hints that simple shape primitives such as “curviness” might activate neural activation and guide this process. Drawing on elementary invariance principles, we propose that a statistical geometric object, the probability distribution of the normalized contour curvatures (NCC) in the intensity field of a planar image, has the potential to represent and classify categories of objects. We show that NCC is sufficient for discriminating between cognitive categories such as animacy, size and type, and demonstrate the robustness of this metric to variation in illumination and viewpoint, consistent with neurobiological constraints and psychological experiments. A generative model for producing artificial images with the observed NCC distributions highlights the key features that our metric captures and just as importantly, those that it does not. More broadly, our study points to the need for statistical geometric approaches to cognition that build in both the statistics and the natural invariances of the sensory world.

The role of semantics in the perceptual organization of shape

Establishing correspondence between objects is fundamental for object constancy, similarity perception and identifying transformations. Previous studies measured point-to-point correspondence between objects before and after rigid and non-rigid shape transformations. However, we can also identify ‘similar parts’ on extremely different objects, such as butterflies and owls or lizards and whales. We measured point-to-point correspondence between such object pairs. In each trial, a dot was placed on the contour of one object, and participants had to place a dot on ‘the corresponding location’ of the other object. Responses show correspondence is established based on similarities between semantic parts (such as head, wings, or legs). We then measured correspondence between ambiguous objects with different labels (e.g., between ‘duck’ and ‘rabbit’ interpretations of the classic ambiguous figure). Despite identical geometries, correspondences were different across the interpretations, based on semantics (e.g., matching ‘Head’ to ‘Head’, ‘Tail’ to ‘Tail’). We present a zero-parameter model based on labeled semantic part data (obtained from a different group of participants) that well explains our data and outperforms an alternative model based on contour curvature. This demonstrates how we establish correspondence between very different objects by evaluating similarity between semantic parts, combining perceptual organization and cognitive processes.

Analisis Morfometrik dan Klasifikasi Bentuk Lutjanus spp. Berdasarkan Gambar Digital

Lutjanus spp is a genus of the Lutjanidae family. The number of Lutjanus spp in waters around the world are 72 species. For this amount, 33 of them living on Indonesian waters. According to the IUCN List (2020), about ten species have decreased in population. One of the causes that population decline in several species is, the recording of capture fisheries has very limited production data. This is caused by the difficulty of identification in the field, which results in the overfishing of certain species. The identification process can be carried out based on morphometric features. Geometric morphometrics can be explaining morphological variations objectively and accurately. There are several methods used to represent the shape of an image in general. Namely point linking, complex coordinate, tangent angle, contour curvature, and triangle-area representation.Lutjanus spp by calculating the value of landmark positions, landmark curvature, changes in landmark angle, landmark distance, and landmark inclination. The results of feature extraction were used to classify Lutjanus spp (Lutjanus argentimaculatus, Lutjanus bohar, Lutjanus carponotatus, Lutjanus fulviflamma, and Lutjanus sebae). The results of this study indicate that the morphometric geometric approach can extract the feature values of the position of landmarks, a curvature of landmarks, changes in the angle of the landmark, distance of landmark, and the inclination of the landmark. The classification results using the Support Vector Machine (SVM) classification technique can distinguish Lutjanus spp with an accuracy rate of 65.03%. Thus, the application of SVM can be used to classify Lutjanus spp species, which will be useful in the identification process. Keywords: clasificasion, identification, morphometric geometric, Lutjanus spp, support vector machine. AbstrakLutjanus spp. adalah salah satu marga dari famili Lutjanidae. Jumlah spesies Lutjanus spp di perairan seluruh dunia yaitu 72 spesies. Dari 72 spesies tersebut 33 diantaranya hidup di perairan Indonesia. Menurut IUCN (2020) sekitar 10 spesies mengalami penurunan populasi. Salah satu penyebab menurunnya populasi pada beberapa spesies yaitu pencatatan data produksi perikanan tangkap masih sangat terbatas. Hal ini disebakan oleh sulitnya identifikasi di lapangan sehingga mengakibatkan overfishing pada spesies tertentu. Proses identifikasi dapat dilakukan berdasarkan ciri morphometrik. Geometri Morfometrik dapat menjelaskan variasi morfologi secara objektif dan akurat. Ada beberapa metode yang digunakan dalam merepresentasi bentuk suatu citra secara umum. yaitu point linking, complex coordinate, tangent angle, contour curvature, serta triangle-area representation. Pendekatan morphometric geometric pada penellitian ini digunakan untuk mengekstraksi fitur bentuk Lutjanus spp. dengan menghitung nilai posisi landmark, kelengkungan landmark, perubahan sudut landmark, jarak landmark, dan kemiringan landmark. Hasil ekstraksi fitur digunakan untuk mengklasifikasikan spesies Lutjanus spp. (Lutjanus argentimaculatus, Lutjanus bohar, Lutjanus carponotatus, Lutjanus fulviflamma, dan Lutjanus sebae). Hasil penelitian ini menunjukkan, bahwa pendekatan Geometri Morfometrik dapat melakukan ekstraksi nilai fitur posisi landmark, kelengkungan landmark, perubahan sudut landmark, jarak landmark, dan kemiringan landmark. Adapun hasil klasifikasi menggunakan teknik klasifikasi Support Vector Machine (SVM) mampu membedakan spesies Lutjanus spp. dengan tingkat akurasi sebesar 65.03%. Dengan demikian, penerapan SVM dapat digunakan untuk melakukan klasifikasi terhadap spesies Lutjanus spp yang akan bermanfaat pada proses identifikasi.Kata kuncis: klasifikasi, identifikasi, geometri morfometrik, spesies lutjanus spp., support vector machine.

P735 Septal curvature - a novel, semi-automated parameter to aid in recognition of basal septal hypertrophy in arterial hypertension

Funding Acknowledgements Horizon 2020 European Commission Project MSCA-ITN-2016 (764738), Grant from Fundacio La Marató de TV3 (040310). Background and aim Localized basal septal hypertrophy (BSH) is a known marker of increased afterload and localized deformation impairment, and can be seen in one-fifth of patients with arterial hypertension. Although there is variability in the classification, BSH is mainly defined from ratios between several wall thickness measurements. We hypothesize that the curvature of the septum is reflective of localized hypertrophy and will be significantly increased in patients with BSH. Speckle tracking endocardial delineations of the left ventricle (LV) can be used to quantify curvature, with the potential to create a novel, semi-automatized parameter for recognition of patients with an increased impact of afterload on cardiac structure and function. Methods An echocardiogram was performed on a total of 149 patients with a diagnosis of long-standing hypertension, treated with at least one antihypertensive drug and on 19 healthy age and sex-matched controls. The interventricular septum thickness was measured at basal and mid-level in the parasternal long axis (PLAX) and 4-chamber (4C) views. BSH was identified from a two-part criterion: both a positive visual assessment of an abrupt change in septal thickness seen in the 4C or PLAX views and a basal to mid-septal ratio ≥ 1.4. A dedicated software for speckle tracking was used to trace the endocardial border of the LV in 4C and 3C view. In post-analysis, we quantified the maximal curvature of the antero- and inferoseptal segments from the exported myocardial contour. Curvature, measured in m-1, was defined as the reciprocal value of the radius of the circle fitted into the curve defined by three subsequent neighboring points in the myocardial contour. Curvature was considered negative if the curve was convex with respect to the LV long-axis. Results Using septal wall thickness measurements, 19% (n = 28) of hypertensive patients were classified as having BSH, whereas all healthy controls had normal geometry. Basal antero- and inferoseptal wall thickness was significantly increased in the BSH group, which was coupled with regional deformation impairment (basal inferoseptum, controls vs. non-BSH vs. BSH: 16.1 ± 2.33 vs. 15.14 ± 2.8 vs. 13.02 ± 2.98 %, p < 0.001). The curvature of the basal inferoseptum was significantly higher in the BSH group (controls vs. non-BSH vs BSH: -23.4 (-27.2, -10.9) vs. -28.3 (-40.2, -19.3) vs. -50.5 (-66.8, -33.9) m-1, p < 0.001) (Figure 1), with the same trend seen in the basal anteroseptum. The inferoseptal curvature showed a moderately strong correlation with the inferoseptal basal-to-mid wall thickness ratio (R = 0.527, p <0.001). Conclusion Increased septal curvature is an easily quantifiable, single-value, semi-automated parameter reflective of localized thickening that could easily be incorporated into the output of the LV speckle tracking workflow, possibly aiding in the recognition of hypertensive patients in need of a closer clinical follow-up. Abstract P735 Figure 1

Spatial variation of channel head curvature in small mountainous watersheds

High-resolution digital elevation models (DEMs) offer opportunities for channel network extraction due to its representation of realistic topography. Channels are generally surrounded by well-defined banks that have a distinct signature in the contour lines. Contour curvature is one of the important topographic attributes usually used for channel head identification; however, the curvature at channel heads may vary considerably between and even within watersheds. Therefore, uncertainty exists in the extracted channel heads due to the specified curvature threshold. In this paper, the locations of channel heads in 14 small mountainous watersheds are obtained using a nonparametric method based on the shape of contour lines generated from DEMs with a spatial resolution of 1 m, and the channel head curvature is computed from the extracted channel heads. The spatial distributions of the channel head curvature in these 14 watersheds have been analyzed, and another two watersheds with field-mapped channel heads are selected for validation. The results indicate that: (1) the channel head curvature is sensitive to the local terrain and varies within and between watersheds; (2) the Gamma distribution effectively fits the spatial distribution of the channel head curvature in all the selected watersheds; and (3) constant threshold-based methods for channel head identification gain significant location errors even within a single watershed.

Mid-level Priming by Completion vs. Mosaic Solutions

We report two experiments on the role of mid-level processes in image segmentation and completion. In the primed matching task of Experiment 1, a cue→prime sequence was presented before the imperative stimulus consisting of target shapes with positive versus negative contour curvature polarity and one versus two axes of mirror symmetry. Priming shapes were included in two composite occlusion displays with the same T-junction information and different geometric features supporting a distinct balance between completion and mosaic solutions. A cue, either congruent or incongruent with targets, preceded the presentation of the composite priming display. Matching performance was affected by primes in the expected direction, while cue congruency participated only in a marginally significant three-way interaction, and prime duration had no effect. In Experiment 2, the cue→prime sequence was replaced by a fixation cross to control for the priming effect obtained in Experiment 1. The study confirmed that contour connectability and curvature polarity are effective structural factors capable of competing with symmetry in mid-level image segmentation and completion processes.

The Spinning Ellipse Speed Illusion

The Spinning Ellipse Speed Illusion is an illusion of perceived speed in which a low-aspect ratio “fat” ellipse will appear to rotate more slowly than a higher-aspect ratio “skinny” ellipse that is rotating at the same speed. This illusory percept can be observed when the ellipses are defined by luminance, color, relative motion, and dotted contours and across a wide range of rotational speeds and eccentricities. The illusion is not limited to rotating ellipses and can be observed with different-shaped contours as well. The Spinning Ellipse Speed Illusion illustrates that the perceived speed of a rotating object depends in part on the form and form features of the object. Objects without characteristic form features such as regions of high or discontinuous contour curvature will appear to rotate more slowly than objects that have these features.

Spectral receptive fields do not explain tuning for boundary curvature in V4

The midlevel visual cortical area V4 in the primate is thought to be critical for the neural representation of visual shape. Several studies agree that V4 neurons respond to contour features, e.g., convexities and concavities along a shape boundary, that are more complex than the oriented segments encoded by neurons in the primary visual cortex. Here we compare two distinct approaches to modeling V4 shape selectivity: one based on a spectral receptive field (SRF) map in the orientation and spatial frequency domain and the other based on a map in an object-centered angular position and contour curvature space. We test the ability of these two characterizations to account for the responses of V4 neurons to a set of parametrically designed two-dimensional shapes recorded previously in the awake macaque. We report two lines of evidence suggesting that the SRF model does not capture the contour sensitivity of V4 neurons. First, the SRF model discards spatial phase information, which is inconsistent with the neuronal data. Second, the amount of variance explained by the SRF model was significantly less than that explained by the contour curvature model. Notably, cells best fit by the curvature model were poorly fit by the SRF model, the latter being appropriate for a subset of V4 neurons that appear to be orientation tuned. These limitations of the SRF model suggest that a full understanding of midlevel shape representation requires more complicated models that preserve phase information and perhaps deal with object segmentation.