A quantitative quality measure approach for 3D photorealistic virtual models has been presented based on various metrics. The approach applied metrics that are used in panoramic images to measure image stitching quality. A new metric has also been proposed to be employed in the presented MTQI approach.

Misalignments in texture mapping due to lens distortions have been managed by developing an automatic distortion management function based on a second-order polynomial equation derived experimentally. The main job of this function is to control misalignments which occur in photorealistic models due to using wide and ultra-wide angled lenses calibrated using self-calibration.

To validate the MTQI approach, the texture quality of three real models has been measured using the proposed approach in cases of ADM OFF and ON. They are also investigated visually by subjects including experienced archaeologists. The results show a consistent correlation.

a) Misalignments; b) Misalignment-free

using the Automatic Distortion Management (ADM) Function

A Photorealistic 3D model for a cubicle inside the Domatilla catacomb

with two zoomed snapshots

As wide angle lenses were employed, the achieved distortion model cannot deliver complete distortion-free images. This causes unavoidable misalignments in case of traditional automatic texture mapping. Accordingly, each model has been textured automatically once when (ADM is OFF) then when (ADM is ON). The images in the models have been captured using different focal distances, while the first model given in the previous section uses a 14 mm lens, the second one uses a 20 mm lens. The third model was generated from two sets of images using both lenses (14mm and 20mm). Figures show three real photorealistic models with zoomed snapshots to show the effect of the ADM function. From the snapshots when (ADM OFF and ON), It is clear that the ADM function succeeded to manage small and large distortions in the three models even when two focal lengths were employed.

The texture quality of the three models is measured using the developed MTQI approach employing the proposed metrics in the case of (ADM OFF and ON). On the other hand, each model has been inspected by several subjects in subjective experiments. The subjects are working in the area of documentation, part of them are in the field of archaeology and architecture. They are asked to assess each model and rank them according to texture quality. The average ranking values, together with the proposed Model Texture Quality Index results, are given in the following Table.

The final photorealistic 3D model (ADM: ON) for the second model

with two zoomed snapshots Left (ADM: OFF) Right (ADM: ON)

The final photorealistic 3D model (ADM: ON) for the third model

with two zoomed snapshots Left (ADM: OFF) Right (ADM: ON)