The paper by T. Gevers and A.W.M. Smeulders from the University of Amsterdam proposes new color models, $c_1c_2c_3$ and $l_1l_2l_3$, that are invariant to viewing direction, object geometry, and shading. The model $l_1l_2l_3$ is also invariant to highlights, and a new photometric color invariant $m_1m_2m_3$ is introduced for matte objects under changing illumination. The performance of these models is evaluated using a database of 500 images of 3-D multicolored objects. The results show that $l_1l_2l_3$ and hue $H$ followed by $c_1c_2c_3$ and normalized colors $rgb$ achieve high object recognition accuracy under white illumination. Only $m_1m_2m_3$ is invariant to changes in illumination color. The paper also discusses the robustness of these models to various factors such as viewpoint, object orientation, illumination intensity, and noise.The paper by T. Gevers and A.W.M. Smeulders from the University of Amsterdam proposes new color models, $c_1c_2c_3$ and $l_1l_2l_3$, that are invariant to viewing direction, object geometry, and shading. The model $l_1l_2l_3$ is also invariant to highlights, and a new photometric color invariant $m_1m_2m_3$ is introduced for matte objects under changing illumination. The performance of these models is evaluated using a database of 500 images of 3-D multicolored objects. The results show that $l_1l_2l_3$ and hue $H$ followed by $c_1c_2c_3$ and normalized colors $rgb$ achieve high object recognition accuracy under white illumination. Only $m_1m_2m_3$ is invariant to changes in illumination color. The paper also discusses the robustness of these models to various factors such as viewpoint, object orientation, illumination intensity, and noise.